8000 feat(voice): read aloud with StyleTTS2 on this machine, as an opt-in service by adeelahmad · Pull Request #291 · adeelahmad/lense · GitHub
[go: up one dir, main page]
Skip to content

feat(voice): read aloud with StyleTTS2 on this machine, as an opt-in service - #291

Merged
adeelahmad merged 1 commit into
mainfrom
claude/project-thread-cy94he
Oct 10, 2026
Merged

feat(voice): read aloud with StyleTTS2 on this machine, as an opt-in service#291
adeelahmad merged 1 commit into
mainfrom
claude/project-thread-cy94he

Conversation

@adeelahmad
@adeelahmad adeelahmad commented Oct 10, 2026
Copy link
Copy Markdown
Owner

Requested by Adeel · project thread

Description

Before: spoken answers and podcasts needed a cloud provider, an OpenAI-compatible server you ran yourself, or a Mac.

After: StyleTTS2 is a text-to-speech choice in Settings → AI assistant → Voice. It runs on your own machine as the optional styletts2 Compose service (docker compose --profile styletts2 up -d), off by default because it is heavy (PyTorch, ~1 GB model fetched on first use). Its voices are clips: drop a few seconds of someone speaking into the styletts2-voices volume (e.g. grandpa.wav) and it appears in the voice picker; default is a public-domain LibriVox reader.

How: styletts2/ is a small stdlib HTTP server around the MIT styletts2 package that speaks the OpenAI audio API (POST /v1/audio/speech, GET /v1/audio/voices, GET /health). Lens's speech.speak() and voices() call it at voice.styletts2_url (default http://styletts2:8880); when it isn't running, answers fall back to the browser and the picker says how to start it. A new StyleTTS2 workflow builds the image and reads a sentence aloud on every change to styletts2/.

Tested for real on CPU (outside Docker, same server and pins): the first request fetched the model and answered in 49 s, later sentences in ~5 s, a clip-cloned voice worked, an unknown voice gave 404.

Motivation and Context

Adeel asked for StyleTTS2 alongside ElevenLabs, macOS say and Gemini (#268).

Types of changes

  • Bug fix (non-breaking change which fixes an issue)
  • New feature (non-breaking change which adds functionality)
  • Breaking change (fix or feature that would cause existing functionality to change; ! in the title)

Checklist:

  • The title is a Conventional Commit.
  • My code follows the code style of this project (make hooks checks it on commit).
  • I added or updated tests.
  • My change requires documentation updates, and I have updated docs/ accordingly.
  • Users need more than the title to understand it, so I wrote it under ## Unreleased in CHANGELOG.md.

Done questions (AGENTS.md)

  • Goal: delivers what the user wanted, merged on green CI.
  • Agents: GET /voice/voices and POST /voice/speak cover it.
  • Zero setup: n/a for now, the service starts with a Compose profile (starting it from Settings is in the DEVLOG refine-later list).
  • Fewest clicks and keystrokes.
  • Configurable: voice.styletts2_url in Settings with a default.
  • Light: off by default; the default install is unchanged.
  • Private and secure: runs locally on the Compose network, no published port; threat model updated.
  • Cost: local, nothing billed; calls go through the activity ledger.
  • User decides: defaults unchanged.
  • Upgrades: new setting has a default; no stored-data change.
  • Honest: docs note the model authors' ask to disclose synthesized speech.
  • Tests, docs and generated files updated.
  • Open standards; dependencies keep Lens MIT (StyleTTS2 code and package are MIT, in a separate image).
  • Result reported with evidence; "refine later" items written down.

🤖 Generated with Claude Code

https://claude.ai/code/session_01WkPdbtd7zvYMRWaRkLN2F6

Generated by Claude Code

@claude
feat(voice): read aloud with StyleTTS2 on this machine, as an opt-in …
9db0215 
…service

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WkPdbtd7zvYMRWaRkLN2F6
@adeelahmad adeelahmad self-assigned this Oct 10, 2026
@adeelahmad
adeelahmad merged commit 4a902a7 into main Oct 10, 2026
16 checks passed
@adeelahmad
adeelahmad deleted the claude/project-thread-cy94he branch October 10, 2026 20:45
@adeelahmad adeelahmad mentioned this pull request Oct 10, 2026
Merged
17 of 19 tasks
adeelahmad added a commit that referenced this pull request Oct 10, 2026
@adeelahmad @claude
docs: list similar content, StyleTTS2 and the Edge clipper as built (#…
58a72fa 
…308)

#296, #291, #285 and #294 merged after the README refresh, so they move from
"In progress" to the feature list, and the web clipper guide's status names
Edge as built.


Claude-Session: https://claude.ai/code/session_01HDx7PPUauyjMn1LZs8JAsS

Co-authored-by: Adeel Ahmad <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

@adeelahmad adeelahmad

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@adeelahmad @claude
0