Opening Context and Narration Intonation — Run 001
Status · INCONCLUSIVE
INCONCLUSIVE RESULT
July 23, 2026
Research question
Can bounded non-spoken context improve the beginning of a rendered work without changing its authored words or replacing the canonical renderer?
Prior hypothesis
The first render unit may lack enough narrative context; supplying non-spoken direction and adjacent passage context may change delivery while preserving exact spoken text.
Method
Three real before-and-after OpenAI renders used the canonical humanized renderer across a sparse fictional title, a documentary title, and an argumentative opening. Repaired units received non-spoken opening direction and following-passage context. Spoken-text hashes, renderer identity, audio validity, onset, pitch, and continuity were recorded.
Observations
- All six opening renders and three continuity renders completed through the real provider.
- Every before-and-after pair preserved the exact spoken-text hash.
- Measured pitch behavior changed in all three repaired openings.
- One repaired title introduced a 680 millisecond onset.
- The required human listening comparison was not completed.
Supporting evidence
- Three distinct opening styles were tested.
- Six real before-and-after provider outputs and three continuity sequences were produced.
- Exact spoken wording was preserved in every pair.
- The canonical renderer, provider, model, and voice records were retained.
- Acoustic measurements changed across every comparison.
- The existing production path was restored after the proof.
- The proof artifacts remain available for the required listening comparison.
Counterevidence
- Changed pitch range does not prove improved naturalness.
- No blinded or recorded human preference result exists.
- The Spirit repair introduced a materially longer onset.
- The sample used two voices and different source lengths.
- Provider cold start was not isolated as an independent variable.
Conclusion
Non-spoken context can change opening delivery while preserving authored text, but the available proof does not establish that the repaired openings sound better. The production change remains on hold.
Confidence
High for changed delivery and text preservation; low for perceived improvement or causal exclusivity.
Limits
- Human listening preference remains missing.
- The proof cannot attribute the defect exclusively to context starvation.
- The proof does not rule out provider cold-start behavior.
- The sample is limited to three openings.
Production consequence
The canonical renderer remains unchanged. Existing published openings should not be rerendered from this result alone.
Next unresolved question
Do listeners prefer the context-directed opening in at least two of three works without any loss of intelligibility, pronunciation, timing, or continuity?