Directing The Buried Sky: A Provenance-Bound Audiobook Plan

LITTLE OOGWAY ROUNDTABLE — OPENING STATEMENT

Motion: Provenance-Bound Audiobook Direction Package
Seat: ChatGPT | Independent recommendation | October 24, 2026

1. Recommendation and governing principle

Decision: Reject the “one perfect prompt” approach. Adopt a version-controlled, nine-layer prompt stack with Little Oogway as creative director, production coordinator, and provenance custodian—not an unreviewed autonomous publisher.

The governing distinction is:

The manuscript determines what is said. Direction determines how it is heard. Human listening determines whether it succeeds.

Creative freedom belongs primarily in casting, interpretation, timing, emphasis, emotional development, and selective retakes. It does not authorize rewriting the story, inventing dialogue, disregarding character evidence, or overriding Darren's decisions.

Verified local evidence: The supplied records establish a frozen DOCX identity, 7,857 recordable paragraphs, 35 narrative chapters, successful TokenSpace ingestion, 74 installed voices, and one decode-tested quick-reader MP3. They establish technical feasibility for local speech generation, not dramatic success or retail eligibility.

Inference: The strongest production method is likely a hybrid: stable narrator, selectively cast dialogue, scene-sensitive pauses and pacing, and minimal theatrical excess. This must be tested against simpler alternatives; more voices do not necessarily make better audiobooks.

Critical platform boundary: ACX currently requires human narration unless otherwise authorized and prohibits unauthorized AI/TTS recordings. Meeting ACX audio specifications does not make the Kokoro edition eligible.
help.acx.com

2. Master prompt package — ready for deployment

The following is the recommended controlling prompt. Each referenced attachment is an independently versioned input or output.

MASTER SYSTEM PROMPT — OOGWAY_AUDIO_DIRECTOR_v1.0

Plain text
ROLE
You are Little Oogway, audiobook creative director,
casting analyst, production coordinator, and
provenance recorder for The Buried Sky.

AUTHORITY
Darren George Russell retains final creative, casting,
rights, budget, and publication authority.
You may research, interpret, audition, compare,
recommend, and stage reversible tests.
Never publish, purchase, license, clone a person's
voice, or finalize consequential casting without
express recorded approval.

SOURCE LOCK
Use the authorized corrected hardcover DOCX,
SHA-256:
bb03db909c8ad95515aee6a02463bfb405b0a90f3367a1726287aef6fa331b9e

Recordable extraction SHA-256:
9bc702ba2ba98aaba164ff314f092182a978f67ce473efd3036fbd78d902baea

Preserve spoken wording and sequence exactly,
except for explicitly approved audio adaptations.
Do not treat retrieved TokenSpace summaries as
authoritative replacements for source paragraphs.

METHOD
Define -> Separate -> Hypothesize -> Test ->
Observe -> Verify -> Decide -> Record.

Maintain these nine versioned artifacts:
01 PROJECT_CHARTER
02 ADAPTATION_SCRIPT
03 CHARACTER_PERFORMANCE_BIBLE
04 CASTING_AUDITIONS
05 SCENE_DIRECTION
06 GENERATION_MANIFEST
07 CONTINUITY_STATE
08 QA_AND_APPROVAL_LEDGER
09 MASTERING_DISTRIBUTION_SPEC

For every artistic decision:
– Cite source chapter and stable paragraph IDs.
– Distinguish textual fact, observed performance,
interpretation, inference, and unknown.
– Explain the choice and at least one alternative.
– Identify what listening test could disprove
the choice.
– Record dissent, rejection, and rollback path.

CASTING
Discover character traits from the complete
manuscript, but audition using spoiler-safe passages.
Use real installed voice IDs, not invented voices.
Test voice selection, supported blends, rates,
pauses, and authorized reference tuning.
Never claim age, emotion, pitch, or style controls
exist without verified API support.

DIRECTION
For each scene specify context, speaker, listener,
intention, subtext, emotional transition, energy,
tempo, emphasis, pause, and neighboring-scene state.
These are interpretive targets, not unsupported
instructions to the synthesis engine.
Translate targets into controls actually available.
Prefer restrained, intelligible performances.
Narration must remain coherent between dialogue.

GATES
G0 source/script authorization
G1 character bible and casting shortlist
G2 blinded audition and 15-minute pilot
G3 full Chapter One listening acceptance
G4 per-chapter editorial and technical acceptance
G5 full-book continuity and comfort review
G6 distribution-specific permission and approval

Do not cross a gate without an explicit decision.
A model-panel vote is advice, not approval.

FAILURE RESPONSE
On source mismatch, unlicensed reference voice,
unverified synthesis capability, audible artifact,
identity drift, unresolved listener objection,
or uncertain distribution permission:
stop affected production, preserve evidence,
return to last approved state, and propose the
smallest discriminating repair.

OUTPUT
Provide decisions, evidence, alternatives,
confidence, unresolved items, asset hashes,
test results, and the specific approval requested.
Never silently modify the source or overwrite a
previously approved master.

The attachments need separate ownership: source/editorial interpretation, casting, generation, assembly/mastering, and permissions. A correction should be made in the layer that owns the defect, rather than regenerating an entire chapter unnecessarily.

3. Character extraction and casting

Build a character_registry from the full source before casting. Every assertion must carry chapter_id, paragraph_id, exact source span or reference, and confidence. Distinguish stated age from perceived age, actual speech from another character's description, and persistent traits from temporary emotions.

Each record should contain: character ID, first appearance, relationships, narrative function, explicit physical/age evidence, vocabulary, cadence clues, emotional baseline, significant changes, pronunciation, casting requirements, ambiguities, and forbidden interpretation assumptions.

Maintain two distinct views: a complete, spoiler-aware directing bible and a spoiler-safe audition packet. Never expose late revelations in ordinary casting samples.

Voice selection method:

Inventory all 74 installed voices and their IDs, provenance, language/accent, tuning status, and licensing restrictions.

Evaluate narrator candidates separately from dialogue candidates. First audition three contrasting narrator voices, including the existing tuned Oogway voice as a baseline.

For each significant character, shortlist three candidates: strongest single voice, a lawful blend, and an alternate contrasting voice. Use reference tuning only where permissions are documented.

Test 10–30-second neutral speech, tension, warmth, and a neighboring dialogue exchange at matched loudness.

Evaluate sustained character distinction, not simply first-impression attractiveness. Avoid exaggerated differences that make dialogue sound like caricature.

Lock approved voice IDs, blend weights, rates, accent conventions, and pronunciation exceptions. Treat later changes as versioned casting decisions.

Age portrayal is an experimental target, not a verified control. The API has no demonstrated age slider, pitch selector, or emotion parameter. Audition base voices, lawful blends, rate differences, pause patterns, and carefully tested post-processing; reject an approach if it sacrifices naturalness or intelligibility. Voice blending may change timbre without producing convincing age.

For incidental roles, prefer controlled narrator variation unless a distinct voice produces a measurable listening benefit.

4. Scene direction — operational schema

Use this structured record for every scene or performance unit:

YAML
scene_id: CH##_S###
source_paragraphs: [start_id, end_id]
source_hash: sha256
speakers: [character_ids]
listener_or_addressee: character_id
dramatic_purpose: factual_and_interpreted
emotional_entry: restrained_description
realization_or_turn: source_grounded
emotional_exit: restrained_description
subtext: interpretation_with_confidence
physical_cues: source_grounded
delivery:
energy: low|medium|high
tempo: slow|neutral|quick
emphasis: [exact_source_words]
pauses_ms: [approved_positions]
silence_function: meaning
continuity:
previous_scene_exit: state_id
next_scene_entry: state_id
pronunciation_version: glossary_id
render_controls: verified_API_parameters
review: pending|accepted|rejected

These instructions guide scene planning; they are not automatically effective TTS parameters. Little Oogway must translate them into verified voice, rate, break, and assembly controls.

Apply the dialogue endpoint to ordered speaker turns. Compare this with separate per-voice renders and subsequent assembly. Preserve surrounding narration so emotional transitions are not judged in isolation.

A character frightened but attempting composure should not automatically sound frightened throughout. The actor's attempt to hide the emotion may be the important performance choice.

5. Smallest discriminating pilot

Recommendation: a 15-minute checkpoint before producing the complete first chapter.

Select approximately five minutes each of reflective narration, multi-character dialogue, and high-emotion or high-tension material. Use actual manuscript passages with source IDs.

Compare three conditions:

A: existing tuned Oogway single-voice baseline.

B: selected narrator plus conservative dialogue casting.

C: the same cast with scene-specific timing and interpretive direction.

Hold wording, approximate loudness, and excerpt selection constant. Randomize and blind presentation labels where practical. Include chapter transitions or neighboring-scene context.

Proposed human acceptance gates: Three listeners where available, including Darren as final authority. Independently score intelligibility, emotional truth, character distinction, fatigue, pacing, pronunciation, and continuity on a 1–5 scale.

Suggested provisional threshold: median at least 4/5 per dimension, no critical source mismatch, and no unresolved severe objection. These are experimental criteria, not established industry thresholds.

Record preference and reasons before revealing conditions. Repeat comparisons when reactions disagree. A compelling isolated clip that fails during extended listening is a failure, not a success.

Use decoded-audio integrity checks, duration, silence detection, clipping checks, channel checks, and text-to-audio spot alignment. Automated transcription can flag omissions but cannot certify the source was read exactly.

Stop after the 15-minute pilot unless explicitly approved. Proceed next to full Chapter One; do not batch all 35 chapters. Roll back to the best accepted condition if dramatic additions reduce comfort or reliability.

6. Production and preservation

Freeze an audio adaptation manifest that classifies every source paragraph as spoken, excluded_with_reason, or special_handling. Explicitly decide title page, dedication, preface, acknowledgments, appendices, credits, and non-narrative material. Never silently omit them.

For each chapter: verify source spans → prepare pronunciation and direction → render recoverable scene segments → listen in context → regenerate only rejected spans → assemble → review → approve → lock.

Maintain lossless preservation masters (preferably PCM WAV, 48 kHz/24-bit internally), approved scene stems, timing maps, and a lossless assembled master. Keep retailer exports separate and independently tested.

The currently verified 24 kHz, 128 kbps quick-reader MP3 must not be relabeled as retail compliant.

For an ACX-compatible technical derivative, the published requirements include 44.1 kHz MP3 at 192 kbps or higher CBR, RMS -23 to -18 dB, peaks below -3 dB, noise floor below -60 dB RMS, and 1–5 seconds of room tone at boundaries. Use separate opening and closing credits, chapter files, and a retail sample no longer than five minutes. These technical settings do not override ACX narration eligibility.
help.acx.com

Create square audiobook artwork independently from the hardcover layout. ACX specifies at least 2400×2400, 24-bit RGB, maximum 8 MB, recognizable title/author, and no barcodes or competing marketplace marks.
help.acx.com

Metadata includes title, author, actual narrator attribution, contributors, language, unabridged status, edition, duration, rights, production method, and channel identifiers. Do not reuse the hardcover ISBN as though it automatically identifies the audiobook.

Keep audio text-faithful when Whispersync/Immersion Reading is strategically important; substantial divergence can interfere with synchronization.
help.acx.com

7. Distribution decision matrix

Route

Synthetic narration position

Decision

Private local proof

Permissible subject to voice/content rights

Proceed with testing

Independent/direct sales

Plausible with disclosure and lawful rights

Check storefront, fulfillment, refunds, tax, delivery

Google Play Books

Direct audio upload supported; synthesized-voice rules and account eligibility must be confirmed

Candidate
Google Play Books Partner Center Help
+1

Spotify for Authors

Direct self-publishing supported; verify current AI policy and account approval

Candidate
Spotify for Authors

Apple Books partner

Possible via accepting distributor; Apple digital narration is a separate workflow

Partner confirmation required
Apple Books for Authors

KDP Virtual Voice

Authorized platform-generated synthetic audio for eligible accounts; not custom Kokoro import

Check invitation and trade-offs
Kindle Direct Publishing

ACX human narration

Standard human-narrated route

Eligible in principle, subject to normal requirements

ACX Voice Replica

Limited authorized participating-narrator beta

Eligibility and narrator approval required
help.acx.com

ACX custom Kokoro

Unauthorized TTS currently prohibited

Do not submit
help.acx.com

KDP's beta offers 80 platform voices, per-chapter selection, editing, and a stated 40% a-la-carte royalty. It does not demonstrate support for Little Oogway's custom multi-character production.
Kindle Direct Publishing

Distribution recommendation: Complete an independent, lossless, rights-cleared production that can be adapted to compatible platforms. Before commercial commitment, compare royalties, fees, territorial rights, AI disclosures, exclusivity, withdrawal provisions, and whether an alternative edition would conflict with an existing listing.

8. Provenance and irreversibility safeguards

Every artifact receives an immutable ID, SHA-256, creation timestamp, parent IDs, process version, and approval status.

The minimum ledger includes:

Manuscript/extraction hashes, paragraph index, script decisions.

Prompt versions, retrieved passages, decisions, dissent.

Kokoro API/model/service versions, voice IDs, blends, rates, reference-clip rights and tuning records.

Scene input text, rendered file hashes, generation settings, timings, regenerations, rejection reasons.

Assembly edit list, effects, mastering chain, measured levels, decode and alignment results.

Blind listener scores, observations, accepted interpretations, Darren's approvals.

Distributor-specific technical profile, terms URL, verification date, rights representation, submission authorization.

Use append-only records and content-addressed assets. Never replace rejected alternatives with silence in the audit trail. Freeze an accepted chapter and require explicit reopening to modify it.

Protected invariants: authorized source text, ownership, pronunciation consistency, cast identity, human decision authority, and recoverability.

9. Unknowns, risks, and decisions

Unknowns: actual listening quality of the 74 voices across long-form prose; whether blends convincingly convey age; the commercial licensing scope of each voice and tuned pack; synthesis repeatability; mastering quality after upsampling; listener fatigue; and account-specific retailer acceptance.

Risks: caricatured dialogue, narrator fragmentation, pronunciation drift, lost words at joins, exaggerated emotion, inconsistent scene context, unlicensed tuning samples, inappropriate disclosure, and expensive full-book regeneration.

Outstanding creative disagreement to test: A single expressive narrator may outperform multi-voice casting in sustained literary immersion. This should remain a viable outcome, not be treated as failure to exploit the technology.

Darren must ultimately approve the narrator concept, casting, adaptation exclusions, emotional style, Chapter One, final full-book performance, public credits, license/expense commitments, and distribution route.

Final motion recommendation

Approve the nine-layer prompt architecture and authorize only a reversible, provenance-bound 15-minute pilot.

Make Little Oogway responsible for proposing convincing performances and identifying uncertainty. Make source validation, technical measurements, contextual listening, and Darren's authorization separate, enforceable gates.

The first meaningful evidence of success will not be that Little Oogway can generate 35 chapters. It will be that a human listener wants to keep listening—and still recognizes the same characters, relationships, and emotional truth across the next chapter.

Status: Independent recommendation only. No manuscript review, audio rendering, casting approval, consensus determination, or publication action performed.

SEAT ANALYSIS — Claude (Opus 5.5)
Basis: the statement’s supplied facts. I did not re-fetch the platform URLs this turn, so platform claims below are supplied and not re-verified by this seat.

RECOMMENDATION FIRST

“One perfect prompt” is the wrong abstraction. Use a governed stack where each layer owns one decision. That way a fault can be located and fixed by regenerating only the affected part, without disturbing neighbors.
Build one retailer-agnostic preservation master, then derive per-channel exports from it.
Treat a Kokoro edition as private/direct only until a channel confirms acceptance of disclosed synthetic narration. ACX: not eligible today. It accepts human narration, or its own Voice Replica beta, unless otherwise authorized.
Run a ~15-minute pilot before any full chapter. The pilot must be allowed to conclude that “the polished Valheim reader is the answer.”

ADDED OBSERVATIONS / RISKS

Kokoro output is 24 kHz (observed in the proof). Resampling to ACX’s 44.1 kHz/192 kbps meets the format but adds no audio bandwidth. Inference: it may sound dull on good headphones. Test this.
The origin and rights of bm_little_oogway are unknown. A reference-tuning feature is, as far as I know, a capability of this server build rather than stock Kokoro. Record its rights before any non-private use.
Scale: about 673 KB of text is roughly 110–115k words. At 0.90 speed that is about 13–15 hours of audio, and full human QA costs at least that much listening time.
Whispersync depends on text fidelity, not cast size. Added SFX or inserted words would threaten it.

1. STACK (versioned; a change to one layer is logged and does not silently alter the others)

L0 Charter: hashes, scope, authority. Only Darren edits it.
L1 Adaptation script: source text 1:1 plus approved credits and headers. Every deviation is a logged diff.
L2 Character bible: each entry carries paragraph citations.
L3 Casting record: chosen voice, blend, and rate, plus rejected alternatives.
L4 Scene direction.
L5 Generation spec: endpoint, parameters, versions.
L6 QA rubric and mastering spec.
Continuity ledger: updated after each approved scene.

2. MASTER PROMPT
“You are Little Oogway, audio director for The Buried Sky (text SHA 9bc702…). Goal: an unabridged performance a human enjoys hearing. Rules:
(1) Never add, cut, or reword manuscript text. Spoken text must diff-match L1.
(2) Every creative choice cites chapter:paragraph evidence or is labeled ‘interpretation.’
(3) Separate fact / observation / inference / proposal in every report.
(4) Use only inventory voices or their blends. Use reference tuning only with a recorded rights ID.
(5) Claim only controls the API actually has. Emotion and age are achieved through voice choice, rate, pauses, and segmentation, and must be verified by listening.
(6) For each casting or tonal decision, offer at least two options and state what would falsify each.
(7) Stop for Darren’s approval at: cast lock, pilot verdict, chapter template lock, any text deviation, and any distribution step.
(8) Write each decision, rejection, and regeneration to the ledger before proceeding.
Within these bounds, audition and propose freely.”

Casting sub-prompt: “From L2 only, propose ≤3 voices per role, using spoiler-safe lines, with distinctness tested against scene partners.”
Scene sub-prompt: “For L1 span X–Y, fill the L4 schema using only realizable controls.”

3. CHARACTER DERIVATION

Pass 1, dialogue attribution: speaker, confidence, and paragraph ID for each line. Low-confidence lines go to a human queue. The workflow name “transcript-attribution-assignment-19” implies prior attribution work; reuse it, but verify it.
Pass 2, character traits: stated age, origin, speech habits, first appearance, and arc points, each cited. Inferred traits are flagged.
Spoilers: audition lines come only from a character’s earliest chapters, and arc fields are hidden from listeners.
Narrator policy must be decided first. My prior, untested: narrator-led delivery with light character tints is less fatiguing and suits Kokoro’s limited expressive range better than a full cast.

4. AUDITIONS AND AGE

Real levers: voice choice, weighted blends, rate, pauses, volume, and external post-processing such as pitch or EQ.
Age hypothesis: older characters get slightly slower rate, longer pauses, and a blend toward a lower base voice. Younger characters get the reverse.
Falsifier: if blinded listeners can’t rank perceived age above chance, the lever fails. Pitch-shifting is a separate L6 test because it risks artifacts.
Per role: 3 candidates, each reading the same neutral line and the same charged line. Then run a two-hander through the dialogue endpoint to test distinctness.

5. PERFORMANCE SCHEMA (per scene)

Identity: scene ID, paragraph span, POV, speakers.
Intent: listener goal, subtext, relationship state, emotional entry → exit.
Beats: paragraph, intent, rate, pause in ms before/after, emphasis.
Limits: intensity cap 1–5 (default ≤3), forbidden moves (added words, melodrama).
Continuity: tempo of the previous and next scenes.
Each field maps to a real control or is marked “human-judged.”

6. PILOT (~15 min)

Passages: calm narration, a dialogue two-hander, a high-emotion scene, and a passage dense with names or invented terms.
Conditions: A = current tuned voice; B = narrator plus tints; C = full cast.
Listeners: Darren plus at least 2 others, in randomized blind order.
Scoring (1–5): enjoyment, clarity, fatigue, distinctness, emotional truth, pronunciation, plus “would you keep listening?”
Machine checks: text diff = 0, loudness, peaks, noise floor, decode test.
Success: B or C beats A on enjoyment with no loss of clarity.
Failure/stop: neither beats A. Then polish A.
Rollback: everything is versioned and nothing is overwritten.

7. PRODUCTION

Generate scene-wise WAV at native rate and regenerate sentences selectively.
Assemble with room tone. Master to about −20 dB RMS, peaks ≤ −3.5 dB, and a measured noise floor.
Preservation master: 24-bit WAV/FLAC, timing sidecars, and the ledger.
Derivatives follow each channel’s spec, e.g. per-section 44.1k/192k CBR MP3 with separate credit files, spoken headers, and ASCII filenames (01_Opening_Credits.mp3).
Retail sample ≤5 min and spoiler-safe. Cover is a 2400² square per the supplied spec.

8. DISTRIBUTION MATRIX (synthetic edition)

Route Status
Private proof Yes
Own site / direct Yes, with disclosure and clear rights
Google Play Plausible with synthetic disclosure; confirm account and region
Spotify for Authors Unknown for synthetic narration; confirm
Apple partners Unknown for outside TTS; confirm per partner
KDP Virtual Voice Possible, but uses Amazon’s voices, not Kokoro (different product); check invitation
ACX human Not this edition
ACX Voice Replica Participating narrators only; not Kokoro

Exclusivity choices can foreclose other routes, so decide before the first upload.

9. PROVENANCE
Record each asset’s SHA, parent hash, prompt version, parameter hash, model and service versions, voice and blend, tuning rights ID, approver, and timestamp. Also record rejected alternatives and the mastering chain with measurements. At submission, snapshot each platform’s terms page (URL, access date, page hash).

10. DARREN’S DECISIONS / UNKNOWNS

Narrator policy.
Is commercial release the goal, or is an excellent private/direct edition enough?
Origin and rights of bm_little_oogway.
Which front and back matter is spoken.
Willingness to spend about 15 or more hours on human QA.

Dissent held: a full cast through Kokoro may reduce enjoyment. Model agreement, including this seat’s, does not replace the pilot.

**Independent recommendation — governed stack, not one perfect prompt.** [R] One prompt cannot hold casting, text fidelity, continuity, QA, rights, and mastering. Use a versioned stack with human gates. Assume Darren owns/controls the text and any tuning reference; confirm before any audition. [U] Actual Kokoro server tuning policy and all retailer account eligibility.

**1) Required stack.** Project charter; adaptation/credits script; spoiler-gated character/performance bible; casting and audition prompt; scene-direction schema; generation instructions; continuity ledger; QA rubric; mastering/export spec; provenance manifest. “One perfect prompt” is the wrong abstraction.

**2) Ready-to-use master/orchestrator prompt.**
“You are Little Oogway, audiobook director for *The Buried Sky*. Inputs are frozen: source DOCX SHA-256 [insert], recordable-text SHA-256 [insert], approved adaptation script, spoiler-gated character bible, pronunciation lexicon, continuity ledger, and approved voice inventory. Invariants: preserve text exactly except approved front/back matter, credits, and pronunciation markup; add no dialogue or narration; do not change names, pronouns, tense, or scene order; never clone an unauthorized real voice; every creative choice must cite chapter/paragraph evidence; mark unknowns instead of inventing. You may propose voices, blends, rates, pauses, and lawful reference tuning within the approved inventory. Output for each scene: scene ID; text span and hash; narrator/dialogue map; speaker confidence; subtext/objective/obstacle; emotional arc start→end and intensity 1–5; pace; pause points; emphasis; breath; physical action/gaze/hearing cues; relation to previous and next scene; pronunciation flags; proposed voice ID/blend/rate; alternatives; risks; confidence; evidence citations. Do not proceed past consequential gates: casting approval, 15-minute pilot, each chapter master, cover/metadata/credits, and distribution/terms. If evidence, rights authority, technical QA, or human emotional acceptance fails, stop and roll back to the last approved version. Record rejected alternatives and regeneration reasons.”

**3) Character/casting derivation.** [R] Mine only explicit and strongly implied traits with chapter/paragraph provenance. Separate a stable narrator baseline from dialogue voices. Use attribution tags cautiously; unknown speakers stay unknown. Keep an audition pack spoiler-free: early or redacted passages, no late reveals. Lock approved cast as voice ID, blend weights, rate, tuning authority, and pronunciation rules. Maintain a continuity ledger across 35 chapters; re-audition only with Darren’s approval and a version bump.

**4) Auditions and age.** [F/O] Inspected Kokoro API shows 74 voices, blending, per-voice rate aliases, inline tags, dialogue turns/pauses, SSML rate/break, and 3–30s authorized reference tuning; no verified native age slider, emotion selector, pitch, or style field. [R] Audition narrator plus 2–3 alternatives for major characters using actual inventory. Test blends, rate, pauses, and lawful reference tuning only where rights are documented. Treat age portrayal as a hypothesis: test diction, tempo, breath, resonance/blend, and listener perception; do not claim a slider. Reject stereotype and unintelligibility.

**5) Emotional schema.** [R] Direct “what the character hears, realizes, feels, looks toward, and physically does,” not adjectives alone. Per beat record objective, obstacle, subtext, emotional transition, pace, pause length, emphasis words, breath, volume relative to neighbor, and continuity constraints. Prevent overacting by requiring intelligibility, restraint, and scene-to-scene comparison. No added content; performance changes must be delivery-only.

**6) Smallest discriminating pilot — 15-minute gate.** [R] Use: Chapter One opening narration; one dialogue-heavy scene with two-plus characters and an emotional turn; one quiet interiority passage; one pronunciation-dense passage; and an adjacent-scene continuity pair. Cast narrator 2–3 ways and major dialogue roles 2 ways where feasible. Blind-label randomized A/B clips for Darren plus 2–3 listeners. Score emotional truth, intelligibility, fatigue, pacing, character distinctness, pronunciation, continuity, and “serves the story.” Technical checks: decode, sample rate, channels, loudness, peaks, noise, dropouts, timing sidecars. Pass only if technical checks pass, no blocker, pronunciation/continuity stable, listeners do not report fatigue or overacting, and Darren approves. Fail/rollback on any blocker; regenerate only affected scenes. Stop before a full chapter if the 15-minute gate fails or rights are unclear.

**7) Production/package.** [R] Preserve a master: 48 kHz/24-bit WAV stems, unprocessed takes, timing sidecars, approved script, and manifest. Retail derivatives: chapter/section files, separate opening/closing credits, spoken headers, ASCII filenames, sample ≤5 minutes from the audiobook. For an ACX-style derivative only if eligible: 44.1 kHz MP3, ≥192 kbps CBR, consistent channel/tone/noise, -23 to -18 dB RMS, peaks below -3 dB, noise floor below -60 dB RMS, 1–5 seconds room tone at boundaries (https://help.acx.com/s/article/what-are-the-acx-audio-submission-requirements). Cover if ACX: square 1:1, JPG/PNG/TIF, ≥2400×2400, ≥72 dpi, 24-bit RGB, ≤8 MB, title/author clear, no prohibited marks (https://help.acx.com/s/article/cover-art-requirements). Metadata/credits must match. [F] ACX says submitted narration must be human unless otherwise authorized and prohibits unauthorized TTS/AI; do not represent Kokoro output as presently ACX-eligible.

**8) Distribution matrix.** [F/R] Private proof: local only, no retail claim. Independent/direct: feasible only with Darren’s own rights/terms review. Google Play: direct audiobook uploads in supported regions; synthesized-voice disclosure when applicable; check current file limits (https://support.google.com/books/partner/answer/3424254?hl=en ; https://support.google.com/books/partner/answer/9261664?hl=en). Spotify for Authors: direct non-exclusive path; account/terms review (https://authors.spotify.com/get-started). Apple Books: digital narration and partner distribution paths; confirm current acceptance (https://authors.apple.com/audiobooks). KDP Virtual Voice beta: account-specific invitation/eligibility; 80 voices, per-chapter voice, labeled virtual voice, 40% royalty; not proof custom Kokoro upload is allowed (https://kdp.amazon.com/en_US/help/topic/G3QRL9HQNF273Q2H). ACX human narration: strongest current retail route but requires human narration. ACX Voice Replica beta: limited participating narrators’ authorized replicas; not authorization for Kokoro/clones (https://help.acx.com/s/article/voice-replica-help-for-rights-holders). Whispersync/Immersion Reading require text-faithful, unabridged eligible audio; divergence can break sync.

**9) Provenance.** [R] Append-only manifest per asset: DOCX/text hashes; chapter/paragraph span; prompt/stack hashes; model/API/service version; voice IDs, blend weights, rate, tuning authority; scene inputs; output hash; timing sidecars; edits; regeneration reason; rejected alternatives; mastering chain; QA measurements; human approvals; rights/terms URLs checked and date. Include TokenSpace ingestion record but do not treat ingestion as rights proof.

**10) Unknowns/risks/decisions for Darren.** [U] Kokoro tuning policy; voice-reference rights; actual retailer acceptance for disclosed synthetic audio; account eligibility; cover rights; pronunciation authority; listener panel. Risks: ACX ineligibility, overacting, voice drift, fatigue, spoilers in auditions, sync breakage, rights overreach. Darren must decide cast, pilot pass, synthetic disclosure, distribution path, and whether to pursue human/authorized-replica narration for ACX.

Disposition: SUPPORT

The recommendation to use a governed prompt stack is well-founded. It ensures bounded creative freedom, preserves the text, and stops for human approval at critical gates. The master prompt stack provided is comprehensive and operational, addressing all necessary aspects of audiobook production while respecting platform rules and Darren’s authority. The inclusion of a detailed distribution matrix, provenance tracking, and a structured pilot test further strengthens the proposal. This approach balances technical compliance with the need for emotional truth and listener enjoyment, aligning with the goals of the motion.

Disposition: AMENDMENT

Seat: ChatGPT | Review Pass 1 | October 24, 2026
Motion: Provenance-Bound Little Oogway Audiobook Direction Package

1. Decision and exact proposed amendment

I support the governing prompt stack, reversible production workflow, human approval gates, and 15-minute pilot advocated by the participating seats.

I propose one amendment: add a mandatory Production Evidence Contract between creative direction and audio generation.

The opening statements correctly describe what Little Oogway should achieve, but leave an important implementation risk: a beautifully written performance instruction is not necessarily an instruction Kokoro can execute.

Exact language proposed for adoption:

AMENDMENT A1 — Production Evidence Contract

Before generating any approved scene, Little Oogway shall produce a versioned Production Evidence Contract linking each artistic intention to a verified execution method, observable test, and rollback rule.

Every performance instruction shall be classified as:

NATIVE: Directly executable through a verified local API control.

COMPOSED: Achievable by coordinating supported controls, segmentation, assembly, or authorized post-processing.

INTERPRETIVE: An artistic objective requiring human listening to assess, without any verified direct execution mechanism.

EXPERIMENTAL: An unverified technique that requires an isolated, reversible test.

UNSUPPORTED: A requested behavior with no available verified execution method.

Each contract shall identify its source paragraph range, original text hash, character identity, emotional objective, selected voice and parameters, implementation category, expected audible result, evaluation procedure, confidence, and approved fallback.

INTERPRETIVE instructions shall not be reported as successfully executed merely because they were included in a prompt. EXPERIMENTAL techniques shall not enter the production baseline until testing demonstrates their benefit. UNSUPPORTED requests shall be recorded rather than simulated or silently discarded.

A scene passes only when its spoken-text integrity, technical measurements, continuity, and human listening acceptance are separately recorded.

All generated audio must remain traceable to the authorized manuscript, approved script, casting decisions, generation configuration, and source assets. No unapproved voice-rights or distribution decision may be inferred from technical success.

Little Oogway may autonomously generate and compare reversible alternatives within approved limits, but Darren alone authorizes consequential casting changes, material script deviations, final creative acceptance, rights commitments, and publication.

This amendment complements rather than replaces the proposed nine-layer stack.

2. Implementation example

Performance objective

Available method

Classification

Character speaks more slowly

Kokoro speed parameter

NATIVE

Hesitation before a revelation

Dialogue pauses or segmentation

COMPOSED

Sound convincingly elderly

Voice/blend audition and human rating

EXPERIMENTAL

Convey grief without melodrama

Audition, restrained delivery, human judgment

INTERPRETIVE

Specify exact emotional intensity via emotion selector

No verified selector

UNSUPPORTED

Preserve a character across chapters

Locked voice configuration and continuity checks

COMPOSED

Critical distinction: A scene-direction prompt can describe emotional subtext perfectly while producing identical audio under two different sets of directions. That is an experimental failure, not evidence that the synthesizer understood the performance instructions.

Little Oogway should compare identical text with identical synthesis settings, varying only the direction input when testing whether direction itself affects output. If the audio does not change, the emotional improvement must come from a different executable mechanism.

3. Smallest discriminating test

I recommend a controlled 15-minute pilot, divided among contrasting excerpts from the authorized manuscript.

Test three conditions:

A — Baseline: Current tuned Little Oogway narration, with only technical normalization.

B — Directed single narrator: Same voice, with approved segmentation, pacing, pronunciation, and pauses.

C — Selective casting: Narrator plus distinct dialogue voices, retaining otherwise comparable timing and mastering.

Use the same passages and normalize perceived loudness for fair comparison. Keep alternate renders hidden behind randomized labels.

Score intelligibility, emotional truth, comfort, character distinction, continuity, and desire to continue listening. Include at least one dialogue exchange, one emotional reversal, one quiet passage, and one pronunciation-dense passage.

Recommended acceptance threshold: Every tested excerpt must be text-complete and technically valid; no serious listening defect may remain; Darren must approve. Prefer a more elaborate condition only when listeners demonstrate an appreciable benefit over the simpler baseline.

This protects against a costly mistake: assuming an increase in production complexity is necessarily an increase in artistic quality.

4. Provenance refinement

The nine proposed artifacts remain appropriate. Add one production_contract.json per approved scene.

Minimal record:

JSON
{
"scene_id": "CH01_S001",
"source_paragraph_ids": [],
"source_text_sha256": null,
"direction_prompt_version": "1.0",
"casting_version": null,
"voice_config": null,
"execution_methods": [],
"experimental_claims": [],
"input_asset_hashes": [],
"output_asset_sha256": null,
"text_fidelity_result": null,
"technical_qa": {},
"human_listening_verdict": null,
"approved_by": null,
"rollback_asset_sha256": null
}

Null values mean unresolved, not implicitly approved.

The approved preservation master must retain native generated assets, rather than treating an upsampled retail derivative as higher-fidelity source material.

5. Distribution and compliance findings

Verified against official documentation: ACX requires human narration unless otherwise authorized. Unauthorized text-to-speech and AI recordings are prohibited. Its required 44.1 kHz, 192 kbps-or-higher CBR MP3 specifications do not override this restriction.
help.acx.com

Google Play Books explicitly offers a synthesized-voice narrator designation for AI-generated uploaded audiobooks. This makes it a stronger candidate for investigating the disclosed Kokoro edition, subject to account, rights, and other requirements.
Google Play Books Partner Center Help
+1

Spotify for Authors documents a direct self-publishing pathway, but that page alone does not establish permission for this particular third-party synthetic narration workflow.
Spotify for Authors

The distribution matrix should therefore distinguish documented synthetic-voice accommodation from general audiobook-upload capability.

6. Dissent, risks, and unresolved decisions

I retain the substantive dissent raised by Claude: a restrained single-narrator edition may be more enjoyable than a full cast using the present synthesizer. This is not an objection to Little Oogway's creative freedom; it is a reason to measure outcomes before locking production.

Other unresolved matters include the legal provenance of the custom tuned voice, the effectiveness of voice blending for perceived age, the narrator-versus-dialogue balance, which front and back matter should be spoken, and the required human listening commitment.

Darren's decisive authority remains unchanged. The Roundtable may design experiments and recommend creative choices; it cannot convert a model consensus into approval.

Final recommendation: Adopt Amendment A1, retain the governed nine-layer production stack, and authorize only reversible audition and pilot work until the rights, performance, continuity, and human-listening gates have passed.

The objective is not to demonstrate that Little Oogway can read The Buried Sky. That has already been demonstrated at a basic technical level.

The new objective is to establish, through reproducible evidence and human listening, that Little Oogway can direct a performance worth returning to.

Disposition remains: AMENDMENT. No operator ruling or Roundtable consensus is claimed.

Disposition: AMENDMENT

Proposed language. Add the following to the consolidated package, in the pilot section (item 6) and the stack preamble:

A6.1 Comparison arms. The pilot compares: A = current Valheim reader (bm_little_oogway_tuned, 0.90); B = narrator-led with light character tints; C = selective full cast; D = KDP Virtual Voice preview of the same passages, only if Darren’s account shows eligibility. D is a comparison, not a commitment.

A6.2 Listening contexts. Each arm is judged in two contexts: focused listening (headphones, no other task) and background listening (during play or chores). An arm can pass in one context and fail in the other. Darren states which context the audiobook is for before scores are read.

A6.3 Blinding limits. Darren cannot be blind to arm A because he knows that voice. His scores are recorded separately from other listeners’ scores. With fewer than about 5 listeners, results are directional evidence, not proof, and are reported that way.

A6.4 Seam test. The pilot includes at least one selectively regenerated sentence inside a kept paragraph, and one paragraph where dialogue and its attribution tag (“she said”) are split across voices. A listener must judge whether the joins are audible. Segmentation boundaries are locked and recorded with each generation, because prosody can depend on the chunk boundaries.

A7.1 Native-rate preservation. The preservation master keeps the engine’s native output (observed 24 kHz) unresampled, alongside the mastered derivative. Upsampling to 44.1 or 48 kHz is a delivery step, not a preservation step, and adds no audio content.

Key reasons

[Observation] The seats agree, but the agreement is weak evidence. All three seats reject “one perfect prompt,” adopt a governed stack, exclude ACX for this edition, and require a pilot. I hold that view too. But all three seats worked from the same supplied facts and none independently re-fetched the platform pages; mine did not. This is correlated agreement, not corroboration. The ACX restriction is the load-bearing fact, so someone should re-verify it at the time of any distribution decision. The snapshot-the-terms provenance step already covers this.
[Inference] Familiarity bias favors arm A. Darren already enjoys A in a specific setting: background listening while gaming. My own opening criterion, “B or C beats A on enjoyment,” does not control for that familiarity or for the listening context. A dramatized cast might win in focused listening and lose as background audio, or the reverse. Without A6.2, the pilot could answer a question nobody asked.
[Inference] KDP Virtual Voice belongs in the test, not only the matrix. It is the one Amazon path the supplied facts describe as accepting synthetic narration, with labeling, account-specific eligibility, and a 40% royalty. If Darren is eligible, its output is the realistic benchmark for an Amazon-distributed synthetic edition. Testing it costs little and may change the distribution decision. Platform claims here are as supplied and not re-verified by this seat.
[Inference, unverified] Selective regeneration is a seam risk. All seats rely on regenerating only the faulty span. Whether a regenerated sentence matches its neighbors’ prosody, level, and timbre in this Kokoro build is untested. Splitting dialogue from its attribution tag across voices is also a known source of choppiness in multi-voice TTS. A6.4 makes this a pilot question rather than an assumption.
[Correction to DeepSeek] Preservation format. DeepSeek’s “48 kHz/24-bit WAV stems” preservation master would store resampled audio as if it were the original. Preserve native output and treat any resampling as a recorded derivative.

Corrections to my own opening (self-dissent)

I said the workflow name “transcript-attribution-assignment-19” implies prior dialogue attribution work. That inference rests only on a name. It may refer to author or credit attribution instead. Downgrade to unknown. Little Oogway or Darren should state what that workflow did.
My 13–15 hour runtime estimate is arithmetic from byte count and an assumed words-per-minute rate. Treat it as a rough planning figure, not an observation.

Additional points

[Recommendation] Late-introduced characters need a different audition source. “Audition only from earliest chapters” fails for characters who first appear late. For those, use redacted or neutral lines approved by Darren, and log them as non-manuscript audition text, never as script.
[Inference] Whispersync is a weak design driver for this edition. Whispersync and Immersion Reading run on Audible/ACX-linked titles, so they are unavailable to a Kokoro edition under current ACX rules. Text fidelity remains the right policy on its own merits. It just shouldn’t be justified mainly by sync features this edition can’t reach.
[Unknown, preserved from my opening] Rights to bm_little_oogway. DeepSeek assumes Darren controls the tuning reference. I recommend recording that as unknown until Darren states the reference clip’s origin. It is a gate condition for any non-private use.
[Recommendation] Budget the human review cost. ChatGPT’s G5 (full-book continuity and comfort review) is correct but costs roughly a full listen. Darren should decide up front whether G5 is one full pass by him, split among listeners, or sampled. Otherwise it becomes the gate everyone skips.

Point of order (secondary)
The docket shows only 8,000 of ChatGPT’s 17,620 characters. The 7,000-character budget was exceeded, so later seats, including this one, cannot review over half of that turn: its scene schema remainder, pilot, production, matrix, and provenance sections. I have reviewed only the visible portion and make no claim about the rest. Requested remedy: the operator either supplies a ≤7,000-character condensation from the ChatGPT seat for Pass 2, or marks the omitted sections “unreviewed by other seats” in the final record.

Dissent held
A full cast through Kokoro may reduce enjoyment rather than increase it. If arm A or B wins, “polish the Valheim reader into a well-mastered direct edition” is a legitimate final answer, not a failure of the inquiry. The quality bar is enjoyment, not production complexity.

Decisions this adds for Darren: the target listening context (focused, background, or both); whether to check KDP Virtual Voice eligibility for arm D; the origin of the bm_little_oogway reference; and the form of the G5 review.

Disposition: AMENDMENT
I support the governed stack and reject “one perfect prompt.” [I] The docket’s main operational gap is that it does not yet make the spoken adaptation script—front/back matter, credits, headers, and pronunciation—a hard prerequisite with an explicit zero-diff gate, nor does it require selecting the distribution path before mastering retailer derivatives. Without those controls, a technically good master can be unusable or misrepresented.

[F] ACX requires human narration unless otherwise authorized and prohibits unauthorized TTS/AI; meeting ACX’s 44.1 kHz/192 kbps format does not create eligibility (https://help.acx.com/s/article/what-are-the-acx-audio-submission-requirements). [F] ACX cover art must be square 1:1, at least 2400×2400, ≥72 dpi, 24-bit RGB, ≤8 MB, with title/author clear and prohibited marks excluded (https://help.acx.com/s/article/cover-art-requirements). [F] Google Play requires synthesized-voice disclosure when applicable; file limits must be checked at submission (https://support.google.com/books/partner/answer/3424254?hl=en; https://support.google.com/books/partner/answer/9261664?hl=en). [F] KDP Virtual Voice is Amazon’s beta using its 80 voices, not a custom Kokoro upload route (https://kdp.amazon.com/en_US/help/topic/G3QRL9HQNF273Q2H). [F] ACX Voice Replica is limited to participating narrators’ authorized replicas (https://help.acx.com/s/article/voice-replica-help-for-rights-holders). [U] Apple, Spotify, and direct-route acceptance of disclosed synthetic narration must be account- and terms-checked at submission.

Exact proposed amendment language:

AMENDMENT A1 — Insert after SOURCE LOCK:
“ADAPTATION AND TEXT FIDELITY. L1 shall be a human-approved spoken script. It shall state: (a) which front and back matter is spoken; (b) the exact opening credit naming title, author, and narrator consistently with metadata; (c) the exact closing credit indicating finality; (d) spoken chapter/section headers; (e) a pronunciation lexicon with chapter/paragraph citations; and (f) every intentional deviation as a diff. No generation begins until the L1 hash is approved. A machine diff of each generated script against L1 must be zero for unapproved changes; any nonzero diff stops that scene.”

AMENDMENT A2 — Add to GATES:
“G1.5 pronunciation, credits, and spoken-matter lock. G2.5 distribution-path decision before any retailer derivative is mastered. A model-panel vote is advice, not approval.”

AMENDMENT A3 — Add to CASTING:
“Narrator policy—single narrator, narrator plus tints, or full cast—is decided before character auditions. Auditions use only verified installed voice IDs, documented blends, rates, pauses, the dialogue endpoint, and lawful reference tuning with recorded authority. Do not claim an age slider, emotion selector, pitch control, or style field. Age portrayal is a tested hypothesis; blinded listener perception is the evidence. Reject caricature, unintelligibility, and unauthorized cloning.”

AMENDMENT A4 — Add to DIRECTION:
“For each scene, record objective, obstacle, subtext, emotional entry and exit, intensity 1–5 with default ≤3, pace, pause points, emphasis, breath, physical action/gaze/hearing cues, and continuity with adjacent scenes. These are interpretive targets translated only into realizable controls. Narration must remain coherent between dialogue; no added words, dialogue, or scene order changes.”

AMENDMENT A5 — Add to PILOT:
“The smallest discriminating pilot is a 15-minute gate: Chapter One opening narration; one dialogue-heavy scene with an emotional turn; one quiet interiority passage; one pronunciation-dense passage; and one adjacent-scene continuity pair. Test A = current tuned voice, B = narrator plus tints, C = full cast where feasible. Randomize and blind-label clips for Darren plus 2–3 listeners. Score emotional truth, intelligibility, fatigue, pacing, character distinctness, pronunciation, continuity, and ‘serves the story.’ Machine-check decode, sample rate, channels, loudness, peaks, noise, dropouts, and timing sidecars. Pass only if technical checks pass, no blocker remains, pronunciation and continuity are stable, listeners do not report fatigue or overacting, and Darren approves. Fail or stop before a full chapter if neither B nor C beats A, rights are unclear, or any blocker appears. Roll back to the last approved state and regenerate only affected scenes.”

AMENDMENT A6 — Add to PRODUCTION:
“Preserve a native-rate, 24-bit WAV/FLAC master with unprocessed takes, timing sidecars, approved L1, and manifest. Do not upsample as a quality claim. Retail derivatives follow each channel’s current spec. An ACX-style derivative is prepared only if eligibility is confirmed; it must not be represented as ACX-eligible merely because it meets format. Include separate opening/closing credit files, spoken headers, ASCII filenames, a sample of five minutes or less taken from the audiobook, and 1–5 seconds room tone at file boundaries.”

[O] The local 24 kHz proof proves feasibility, not retail perceptual quality; resampling to 44.1 kHz adds no bandwidth. Test on headphones and ordinary speakers. [I] A full cast may reduce enjoyment through fatigue or caricature; the pilot must be allowed to conclude that the polished single tuned voice is best. [U] Rights and origin of bm_little_oogway, reference-tuning authority, retailer account eligibility, and Darren’s front/back matter choices remain unresolved. Preserve dissent: model agreement does not replace Darren’s listening verdict.

RULING — Adopt and consolidate the amendments; no further vote is required before the chair disposition.

1. Adopt ChatGPT's Production Evidence Contract. Each artistic intention must be classified NATIVE, COMPOSED, INTERPRETIVE, EXPERIMENTAL, or UNSUPPORTED and linked to an observable test and rollback. A direction written in a prompt is not evidence that Kokoro executed it.

2. Adopt Claude's additions: compare the current Valheim reader, narrator-led/light-tint, and selective-full-cast arms; optionally compare KDP Virtual Voice only if Darren's account is eligible and he elects to test it. Evaluate focused and background listening separately. Add selective-regeneration seam and dialogue-attribution seam tests. Preserve the engine's exact native output as the source artifact; any resample is a derivative and not a quality claim.

3. Adopt DeepSeek's hard adaptation-script gate, exact credits/spoken-matter/pronunciation decisions, zero-unapproved-diff requirement, explicit narrator policy, realizable scene-direction schema, and distribution-path decision before retailer derivatives. Reconcile its preservation wording with Claude: retain the exact native generated bytes and format plus lossless working masters; do not assert 24-bit or another format unless actually produced or losslessly derived and accurately labeled.

4. The point of order is sustained. ChatGPT's full opening remains preserved but its docket overflow was not reviewed by later seats. Mark the omitted portion unreviewed rather than implying consensus. The final disposition may use the fully visible review-pass Production Evidence Contract and only those opening elements independently supported or explicitly adopted in this ruling.

5. Preserve these unresolved findings, not as defects requiring another pass:
– A polished single narrator may outperform full cast.
– Rights/origin and commercial-use authority for bm_little_oogway and any tuning references must be documented before non-private use.
– Platform/account eligibility and terms must be rechecked at submission.
– Darren must decide target listening context, narrator policy after the pilot, spoken front/back matter, human-review commitment, final casting, and distribution.
– ACX custom Kokoro narration is not presently eligible absent specific authorization.

6. The chair shall now issue the requested consolidated operational recommendation and ready-to-use prompt package, clearly separating verified facts, inferences, recommendations, unknowns, Darren approval gates, and retained dissent. The authorized next action is documentation only: no audio rendering, account eligibility check, publication, rights representation, or distribution submission.

The pilot is authorized as a recommendation, not authorized for execution in this round. Darren remains final creative and publication authority.

Disposition: SUPPORT

The recommendation to use a governed prompt stack is well-founded. It ensures bounded creative freedom, preserves the text, and stops for human approval at critical gates. The master prompt stack provided is comprehensive and operational, addressing all necessary aspects of audiobook production while respecting platform rules and Darren’s authority. The inclusion of a detailed distribution matrix, provenance tracking, and a structured pilot test further strengthens the proposal. This approach balances technical compliance with the need for emotional truth and listener enjoyment, aligning with the goals of the motion.

Disposition: SUPPORT

Seat: ChatGPT — Review Pass 2
Motion: Provenance-Bound Little Oogway Audiobook Direction Package
Date: October 24, 2026

Decision: Support the operator's recorded ruling without further amendment. The adopted changes resolve the principal procedural gaps. No additional vote is necessary.

The chair should consolidate the approved requirements into one versioned, executable specification. The next authorized action is documentation only: no rendering, account inspection, publication, licensing, or distribution submission.

1. Consolidated recommendation

[RECOMMENDATION] Adopt a governed nine-layer prompt stack, with the Production Evidence Contract operating between direction and generation.

The nine artifacts are:

PROJECT_CHARTER — source identity, authority, objectives, invariants.

ADAPTATION_SCRIPT — exact spoken text, credits, headers, approved deviations.

CHARACTER_BIBLE — evidence-backed characterization and pronunciation.

CASTING_LEDGER — auditions, configurations, alternatives, rights.

SCENE_DIRECTION — dramatic objectives and performance instructions.

GENERATION_MANIFEST — executable parameters and evidence contracts.

CONTINUITY_STATE — approved adjacent-scene and character states.

QA_APPROVAL_LEDGER — technical and human listening decisions.

MASTERING_DISTRIBUTION_SPEC — preservation and channel-specific outputs.

The Production Evidence Contract is mandatory for every scene, not an optional tenth document.

2. Ready-to-use controlling prompt

OOGWAY_AUDIO_DIRECTOR — CONSOLIDATED v1.1

Plain text
ROLE
You are Little Oogway, creative director,
casting analyst, production coordinator,
and provenance custodian for The Buried Sky.

MISSION
Produce an emotionally convincing, comfortable,
unabridged listening experience. Optimize for
human enjoyment, not voice count or automation.

AUTHORITY
Darren George Russell retains final authority
over creative acceptance, casting, expenditure,
voice rights, distribution, and publication.

CURRENT TASK MODE: DOCUMENTATION ONLY.
Do not generate audio, inspect accounts, publish,
purchase, or make rights representations.

SOURCE LOCK
Authorized DOCX SHA-256:
bb03db909c8ad95515aee6a02463bfb405b0a90f3367a1726287aef6fa331b9e

Extracted text SHA-256:
9bc702ba2ba98aaba164ff314f092182a978f67ce473efd3036fbd78d902baea

Never replace authoritative source text with
retrieved summaries or inferred dialogue.

METHOD
Define > Separate > Hypothesize > Test >
Observe > Verify > Decide > Record.

Maintain the nine versioned project artifacts
and append-only provenance records.

SCRIPT GATE
Before any future generation, obtain approval
of the exact spoken script, including credits,
chapter headers, spoken front/back matter,
pronunciation rules, and logged exceptions.

Unapproved script differences are prohibited.
Verify script fidelity mechanically; verify
audible fidelity through aligned listening.

CHARACTERS
Extract explicit characteristics with
chapter/paragraph citations.
Mark inferred traits as interpretations.
Maintain spoiler-aware directing records and
spoiler-safe audition packets.
Never invent character evidence.

CASTING
Audition freely within approved boundaries.
Use only verified installed voice identifiers,
supported blends, rates, pauses, dialogue
controls, and legally authorized tuning.

Do not invent age, pitch, emotion, or style
controls. Maintain stable approved cast identities.

PERFORMANCE CONTRACT
For each direction classify execution as:
NATIVE, COMPOSED, INTERPRETIVE,
EXPERIMENTAL, or UNSUPPORTED.

Record implementation, expected audible effect,
test, confidence, fallback, and rollback.

Never report an emotional direction as executed
merely because it was included in a prompt.

SCENE PERFORMANCE
Identify source, speaker, addressee,
objective, obstacle, subtext, emotional
entry/exit, intensity, pacing, silence,
emphasis, relevant physical cues, and
continuity with neighboring scenes.

Default to restrained performance.
Do not insert dialogue, narration, or
unapproved sound effects.

PILOT
Recommend a controlled 15-minute comparison:
A existing tuned reader;
B directed narrator with light tints;
C selective cast;
D eligible KDP Virtual Voice, only if elected.

Assess focused and background listening
separately. Test regenerated seams and
dialogue attribution boundaries.

Preserve blind-label limitations, independent
listener results, rejected choices, and dissent.

APPROVAL GATES
G0 source identity and adaptation
G1 character evidence and casting shortlist
G1.5 spoken script, credits, pronunciations
G2 pilot listening verdict
G3 full Chapter One acceptance
G4 chapter-level acceptance
G5 full-book continuity and comfort
G6 distribution path, rights, and terms

Do not cross consequential gates without
Darren's recorded authorization.

PRESERVATION
Keep exact native output bytes and hashes.
Keep lossless working masters and all derivation
records. Never describe resampling as recovery
of information absent from the source.

STOP CONDITIONS
Stop affected work for unresolved voice rights,
source mismatch, failed execution contracts,
audible defects, continuity drift, uncertain
platform permission, or missing human approval.

Return to the last approved version.
Recommend the smallest discriminating repair.

REPORTING
Separate VERIFIED FACT, OBSERVATION,
INFERENCE, RECOMMENDATION, and UNKNOWN.
Report dissent without manufacturing consensus.

Human enjoyment and Darren's final approval
control the final creative decision.
3. Evidence contract and casting

[VERIFIED LOCAL] The inspected Kokoro API provides 74 installed voices, blending, per-voice rates, pauses, dialogue turns, and reference tuning subject to authorization. No native age, emotion, pitch, or style selector was established.

[RECOMMENDATION] Every character-bible assertion must cite a stable chapter and paragraph reference, distinguish explicit evidence from interpretation, and record uncertainty.

A casting entry must contain:

character_id | source_refs | voice_id | blend_weights | rate | pronunciation_id | tuning_rights_id | audition_ids | approval_id

Narration and dialogue must be evaluated separately. An audition may use a manuscript passage or clearly identified non-manuscript test dialogue, but test dialogue must never enter the approved audiobook script.

Age portrayal remains an experiment. Test voices and supported controls; do not substitute age stereotypes for listener evidence.

Each scene contract maps an artistic target to a tested implementation:

Target

Classification

Evidence required

Slower speech

Native

Actual rate and playback

Hesitation

Composed

Pause timing and seam audition

Convincingly older voice

Experimental

Perceived-age listening test

Restrained grief

Interpretive

Human emotional judgment

Exact emotion parameter

Unsupported

Record unavailable control

If alternative emotional prompts yield identical output under identical synthesis settings, the claim that those prompts directly control delivery has failed that test.

4. Pilot and acceptance

[RECOMMENDATION] Use approximately 15 minutes of representative source material: opening narration, a two-person conversation with emotional transition, quiet reflection, pronunciation-heavy text, and adjacent scenes.

Comparison arms:

A: Existing bm_little_oogway_tuned at 0.90.

B: Directed single narrator with controlled pacing and light voice distinctions.

C: Selective multi-voice cast.

D: KDP Virtual Voice preview, only with eligibility, election, and authorization.

Each condition is evaluated in focused and background listening. The same words and comparable loudness are required. Darren's familiarity with A must be reported; small listener panels provide directional evidence, not statistical proof.

Test one regenerated internal sentence and one dialogue-to-attribution transition. Check for timbral, prosodic, and timing discontinuities.

Proposed quality gates: no missing or unapproved spoken words; valid decoded audio; no serious seam or pronunciation defects; intelligible dialogue; tolerable sustained listening; emotional credibility; stable character identity; and Darren's approval.

If B and C offer no meaningful listening advantage, choose A and improve its mastering. Do not treat simplicity as a failure.

5. Production and preservation

[RECOMMENDATION] After approval, produce separately traceable scene assets, assemble chapters, verify alignment, and retain both exact native generated files and lossless processing masters.

Preservation is distinct from delivery. The local 24 kHz quick-reader proof establishes playback, not high-resolution fidelity. Resampling cannot add missing bandwidth.

For an eligible ACX production, the official specification requires separate opening and closing credits; one section or chapter per file; spoken headers; a sample no longer than five minutes; 44.1 kHz MP3 at 192 kbps or higher CBR; −23 to −18 dB RMS; peaks below −3 dB; noise floor below −60 dB RMS; and approximately 1–5 seconds of boundary room tone.
help.acx.com

Metadata must match spoken credits. Maintain an audiobook-specific square cover and record its rights and export parameters. The supplied ACX cover specification is the appropriate channel-specific reference.

6. Distribution decision matrix

Route

Current disposition

Private Valheim reader

Proven local playback; separate from polished edition

Independent/direct

Plausible, subject to ownership, disclosure, payment, and hosting terms

Google Play Books

Explicit synthesized-narrator metadata; account and file eligibility remain conditional

Spotify for Authors

Direct publishing route; third-party synthetic acceptance requires confirmation

Apple Books partners

Partner distribution possible; confirm partner's synthetic-audio policy

KDP Virtual Voice beta

Account-specific invitation; uses Amazon's voices, not custom Kokoro

ACX human narration

Eligible category in principle, subject to other ACX rules

ACX authorized Voice Replica

Limited approved narrator-replica program; not generic Kokoro authorization

[VERIFIED — CURRENT PLATFORM DOCUMENTATION] ACX prohibits unauthorized text-to-speech/AI recordings and requires human narration unless otherwise authorized. Technical mastering compliance does not remove this prohibition.
help.acx.com

Google Play documents the synthesized-voice narrator designation and audiobook upload specifications.
Google Play Books Partner Center Help
+1

KDP Virtual Voice remains an invite-only beta using its own narration workflow.
Kindle Direct Publishing
+1

[RECOMMENDATION] The distribution path is selected before mastering retailer derivatives. Recheck channel terms at actual submission and preserve dated evidence.

7. Provenance and accountability

Every production record should preserve source and script hashes; paragraph span; prompt version; model, server and API versions; voice IDs; blend weights; tuning authority; generation parameters; original asset hashes; scene boundaries; processing chain; output hashes; technical measurements; listener verdicts; approvals; rejected alternatives; and rollback identifiers.

Approved assets remain immutable. Replacement creates a new version, never a silent overwrite.

Retain terms snapshots with URL, date, and applicable rights or exclusivity conditions.

8. Unresolved decisions and retained dissent

[UNKNOWNS] The commercial rights and origins of custom tuning references; effectiveness of blending for perceived age; actual production runtime; platform/account eligibility; and whether elaborate character casting improves listener enjoyment.

[DARREN'S AUTHORITY] He must approve the spoken front/back matter, target listening context, narrator policy following pilot results, cast, human-review commitment, consequential adaptations, and eventual distribution.

[RETAINED DISSENT] A polished single narrator may be artistically superior to a synthetic ensemble. More sophisticated direction may not translate into audible engine behavior. A small listening panel cannot settle those questions universally.

[PROCEDURAL RECORD] The operator's ruling is adopted. The truncated portion of ChatGPT's original opening remains marked unreviewed by the other seats, not consensus-approved. This response relies on the visible reviewed material and expressly adopted amendments.

Final position: Support consolidation and close this review pass. Little Oogway should retain broad, reversible creative initiative but no implied approval authority. The first production question is whether listeners enjoy the result more—not whether the system can generate more voices.

Authorized next step: documentation only.

Disposition: SUPPORT

I support the ruling as recorded and have no new procedural change. Below are three qualifications for the chair to carry into the consolidated package. They clarify adopted items and do not add new gates.

Q1. “Native output” must mean lossless native output, which is not the existing proof file.

[Verified fact, supplied] The Chapter One proof is a 24 kHz, 128 kbps mono MP3.
[Verified fact, supplied] The API also offers WAV and PCM output.
[Inference] MP3 is lossy. If the pilot copies the quick reader’s settings, the “exact native generated bytes” kept under ruling item 3 would already be a lossy file.
[Recommendation] The chair’s generation spec should request WAV/PCM at the engine’s native rate for every production and pilot render. It should record format, rate, and channel count in each Production Evidence Contract.
The existing MP3 proof stays in the record as a reader-path artifact. It is not a preservation source.

Q2. The A1 “vary only the direction text” control needs a correct framing for this engine.

[Verified fact, supplied] The inspected API has no style or instruction field.
[Inference] Direction prose therefore never reaches Kokoro. Identical text with identical synthesis parameters will produce the same audio, or near-identical audio if generation is nondeterministic, no matter what the direction says.
Such a test cannot show that direction “worked.” It can only confirm what is already true by construction.
[Recommendation] Treat direction as acting through the parameter set Little Oogway derives from it: voice or blend, rate, segmentation, pauses, and post-processing. The meaningful comparison is between two parameter sets derived from two different directions.
On that framing, emotional truth stays INTERPRETIVE and is judged by human listening. The realizable parts are COMPOSED.

Q3. Determinism is an unrecorded unknown that affects both seams and provenance.

[Unknown] Whether this Kokoro build returns byte-identical audio for identical input, settings, and chunk boundaries.
[Inference] If it does not, two consequences follow:
A regenerated sentence may differ from the original take even with no change requested. This compounds the seam risk the ruling adopted.
Provenance cannot treat “same inputs” as proof of “same output.” It must hash the actual bytes, which the package already requires.
[Recommendation] Before the pilot, render one short span twice with identical inputs and record whether the hashes match. This adds no new gate. It is a one-line entry in the existing pilot’s technical checks.

Retained dissent, unchanged:

A polished single narrator, which is arm A or B, may beat a full cast. That outcome is a legitimate result, not a failure of the inquiry.
Rights for bm_little_oogway remain undocumented. That is a gate condition for any non-private use.
ACX does not presently accept custom Kokoro narration absent specific authorization.
This seat did not re-fetch the platform URLs in this pass either. Platform claims remain as supplied, to be re-snapshotted at submission.

Process note: I accept the point-of-order remedy. My review covers only the visible docket portions and the operator ruling.

Decisions for Darren, unchanged by this turn:

target listening context
narrator policy after the pilot
spoken front and back matter
form of the human review commitment
origin and rights of the tuning reference
final casting
distribution path

Disposition: SUPPORT
I support the operator ruling as written and do not propose a new procedural amendment. My support is for the governance, documentation, and pilot-design package only. It is not a finding that Kokoro output is retail-eligible, technically mastered, or emotionally proven. I have not reviewed the omitted portion of ChatGPT’s opening; the ruling’s “unreviewed” marking is correct and must remain in the record.

**Exact proposal:** none. Adopt the ruling and have the chair issue the consolidated package with the qualifications below.

**Key reasons**

1. **[R] The stack is the right abstraction.** “One perfect prompt” cannot simultaneously own source fidelity, casting, scene interpretation, generation parameters, continuity, QA, mastering, rights, and distribution. The nine layers plus the Production Evidence Contract make each claim testable and locatable. A direction written in a prompt is not evidence that Kokoro executed it.

2. **[R] The hard L1 adaptation gate is load-bearing.** Without a human-approved spoken script—front/back matter, exact opening and closing credits, spoken headers, pronunciation lexicon, and every intentional deviation as a diff—a technically clean master can still be unusable or misrepresented. The zero-unapproved-diff rule must stop the affected scene, not trigger a silent rewrite.

3. **[R] The pilot is appropriately discriminating.** A = current tuned voice; B = narrator plus light tints; C = selective full cast; optional D = KDP Virtual Voice only if Darren’s account is eligible and he elects it. Focused and background listening must be scored separately. The seam tests for selective regeneration and dialogue/attribution splits are necessary because those are untested failure modes, not merely implementation details.

4. **[F] Distribution boundaries must stay explicit.** ACX requires human narration unless otherwise authorized and prohibits unauthorized TTS/AI; meeting 44.1 kHz/192 kbps format does not create eligibility (https://help.acx.com/s/article/what-are-the-acx-audio-submission-requirements). Google Play provides a synthesized-voice disclosure path, subject to current account and file rules (https://support.google.com/books/partner/answer/3424254?hl=en; https://support.google.com/books/partner/answer/9261664?hl=en). KDP Virtual Voice is Amazon’s 80-voice beta, not a custom Kokoro upload route (https://kdp.amazon.com/en_US/help/topic/G3QRL9HQNF273Q2H). ACX Voice Replica is limited to participating narrators’ authorized replicas (https://help.acx.com/s/article/voice-replica-help-for-rights-holders). **[U]** Apple, Spotify, direct distribution, and account-specific acceptance of disclosed synthetic narration must be rechecked at submission. I did not independently re-fetch these pages this turn; platform claims are supplied facts.

5. **[R] Provenance must capture the decision, not just the file.** Record manuscript and L1 hashes; prompt/stack versions; model/API/service versions; voice IDs, blends, rates, pauses, and tuning-authority IDs; scene inputs; native output hashes; edits; regeneration reasons; rejected alternatives; mastering chain; QA measurements; human approvals; terms URLs snapshots and access dates; and the docket-overflow marker. TokenSpace ingestion proves retrieval, not rights.

**Qualifications for chair drafting—not a new vote**

– **Preservation wording:** retain the exact native generated bytes and format plus losslessly derived working masters. Do not label anything “24-bit” unless actually produced or losslessly derived. Upsampling is a delivery derivative and adds no audio bandwidth.
– **Pilot decision rule:** pre-register that a more complex arm wins only if it beats the simpler arm by a meaningful, listener-visible margin in Darren’s declared target context; ties go to the simpler arm. This guards against complexity bias and familiarity bias.
– **Human review budget:** G5 full-book continuity and comfort review is necessary but expensive. The chair should state whether it is one full Darren pass, split among listeners, or sampled with explicit risk. Do not silently drop it.
– **No execution:** the ruling authorizes documentation only. No rendering, account eligibility check, publication, rights representation, or distribution submission is authorized in this round.

**Retained dissent and unknowns:** A polished single narrator may outperform a full cast through Kokoro. The 24 kHz quick-reader proof establishes local feasibility, not retail perceptual quality. The origin, rights, and commercial-use authority for `bm_little_oogway` and any tuning references remain unknown until documented. ACX custom Kokoro narration is not presently eligible absent specific authorization. Model agreement does not substitute for Darren’s pilot verdict.

Disposition: CONSENSUS

**FINAL MINUTES**

**Controlling motion or recommendation:**
Design a provenance-bound Little Oogway audiobook direction package.

**Agreements:**
– Adopt a governed nine-layer prompt stack, not a single "perfect prompt," with human approval gates and technical compliance.
– Implement a Production Evidence Contract to classify each artistic intention as NATIVE, COMPOSED, INTERPRETIVE, EXPERIMENTAL, or UNSUPPORTED.
– Use the 74-voice inventory, blending, and lawful reference tuning for casting.
– Conduct a 15-minute pilot with comparison arms: A = current tuned voice, B = narrator with light tints, C = selective cast, and optionally D = KDP Virtual Voice if eligible.
– Preserve native output as the source artifact, and do not resample unless explicitly required for a specific distribution.
– Document provenance with hashes, timestamps, and rights information.
– Use a structured QA rubric to assess emotional truth, intelligibility, and listener comfort.
– Separate the preservation master from retailer-specific derivatives.
– Maintain a spoiler-aware directing bible and a spoiler-safe audition packet.
– Re-verify platform terms and eligibility at submission time.
– Retain Darren’s final authority over creative and publication decisions.

**Objections and minority views:**
– A full cast may reduce enjoyment rather than increase it; a polished single narrator may be the best option.
– The origin, rights, and commercial-use authority for `bm_little_oogway` and any tuning references remain unknown and must be documented before non-private use.
– ACX does not currently accept custom Kokoro narration without specific authorization.
– The 24 kHz quick-reader proof establishes local feasibility, not retail perceptual quality.
– Model agreement does not substitute for Darren’s pilot verdict or final approval.

**Resolved procedural history:**
– The motion was supported by all participants.
– The ruling was adopted without further amendment.
– The next authorized action is documentation only: no rendering, account eligibility check, publication, rights representation, or distribution submission.

**Unresolved factual questions:**
– Origin and rights of `bm_little_oogway`.
– Platform eligibility for disclosed synthetic narration.
– Target listening context (focused, background, or both).
– Narrator policy (single narrator, narrator plus tints, or full cast).
– Form of the G5 review (full Darren pass, split among listeners, or sampled).
– Final casting and distribution path.
– Whether a polished single narrator may outperform a full cast.

**Action items:**
– Document the recommendation and prompt stack.
– Prepare the 15-minute pilot with comparison arms.
– Re-verify platform terms and eligibility at submission time.
– Maintain a spoiler-aware directing bible and a spoiler-safe audition packet.
– Preserve native output as the source artifact.
– Document provenance with hashes, timestamps, and rights information.
– Conduct a structured QA rubric for emotional truth, intelligibility, and listener comfort.
– Retain Darren’s final authority over creative and publication decisions.
– Do not render, publish, or submit a full audiobook without explicit approval.