How long can one Gemini 3.8 TTS request get before a fresh start inside it?

Open
#324 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
43/100
Issue type
Feature
Clarity
Mostly clear
Activity status
Active
Tech stack
typescript

Research direction

Read the probe archive on the research/gemini-long-form-narration-seams branch and issue #319, then build a listening-page prototype on a prototype/<name> branch. Synthesize the named fixture articles with Charon at ×1.05 narration pace in chunks of rising lengths, and compare with per-unit overlap and cut. Done means reporting the largest length without a fresh start, or that none beats the current approach, and noting whether restarts occur at paragraph pauses or mid-sentence.

Written by the indexing model from the issue text.

Description

wayfinder:prototype

Question

How long can one Gemini 3.8 Flash-Lite TTS request get before a listener hears a fresh start inside it?

Users report fresh starts inside one long request, tied to internal pieces of about 290 words, and a 1,003-word probe showed speech-level plateaus about every 300 words but was never listened to (How do others make long-form Gemini narration flow across segment boundaries?; its probe archive is on the branch research/gemini-long-form-narration-seams). Build a listening page that synthesizes fixture articles (Anthropic, Cloudflare, derStandard, as in the earlier maps' listening pages) as narration chunks of rising length, for example one narration block, about 200, 290, 500 and 1,000 words, with the Charon voice and the speech policy's ×1.05 narration pace, and let the operator judge by ear where a fresh start appears inside one request, compared with speech policy 6's per-unit chunks. The continuity bar is what overlap and cut reached before it was withdrawn on 2026-11-13 (ADR 0028): flow at 9 of 10 boundaries, against 4 of 10 for policy 6 (#291).

The answer is the largest narration chunk that stays free of a fresh start, or the finding that none larger than today's beats policy 6's per-unit chunks. Note whether a fresh start falls at a paragraph pause, where a listener might accept it, or mid-sentence. Prototype material goes on a prototype/<name> branch.

Part of Map: decouple speech synthesis from synchronization units.

Dominant language
TypeScript
Stars
1
Forks
0
PR merge metrics
No merged PRs in 30d

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from patricktree/cup

All issues in patricktree/cup

Similar issues

More TypeScript issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.