Mercurial

comparison dictation/README.md @ 282:6bd4a3990696 default tip

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add cross-platform canvas orchestration and typed composer Enable native ARM dependencies, macOS Copilot orchestration, portable service supervision, CPU dictation, and a shared N-key speak-or-type composer. Co-authored-by: Copilot <[email protected]> Copilot-Session: f68442b1-fa8f-46a0-9689-81710613bbd4
author MrJuneJune <me@mrjunejune.com>
date Thu, 20 Aug 2026 20:45:25 -0700
parents 49e9e591c9bb
children
comparison
equal deleted inserted replaced
281:c57149ad216e 282:6bd4a3990696
1 # WebRTC dictation 1 # WebRTC dictation
2 2
3 This Bazel package runs a local WebRTC speech-to-text service using aiortc and 3 This Bazel package runs a local WebRTC speech-to-text service using aiortc and
4 faster-whisper. The default multilingual Whisper `small` model uses the RTX 4 faster-whisper. The default multilingual Whisper `small` model uses CUDA with
5 4070 Ti through native WSL CUDA with `int8_float16` compute. 5 `int8_float16` on Linux and CPU inference with `int8` on macOS.
6 6
7 ## Setup 7 ## Setup
8 8
9 ```bash 9 ```bash
10 bazel run //dictation:preflight 10 bazel run //dictation:preflight
28 bazel run //dictation:webrtc_smoke -- /path/to/audio.wav 28 bazel run //dictation:webrtc_smoke -- /path/to/audio.wav
29 ``` 29 ```
30 30
31 The model is stored under `~/.cache/zenbu/faster-whisper-small` by default. 31 The model is stored under `~/.cache/zenbu/faster-whisper-small` by default.
32 Override it with `DICTATION_MODEL_DIR`. 32 Override it with `DICTATION_MODEL_DIR`.
33 Set `DICTATION_DEVICE=cpu|cuda` to override platform selection.
33 34
34 Useful tuning: 35 Useful tuning:
35 36
36 ```bash 37 ```bash
37 DICTATION_COMPUTE_TYPE=int8_float16 \ 38 DICTATION_COMPUTE_TYPE=int8_float16 \
44 The first version intentionally binds to loopback and does not configure STUN 45 The first version intentionally binds to loopback and does not configure STUN
45 or TURN. Remote WebRTC access requires HTTPS, authentication, origin policy, 46 or TURN. Remote WebRTC access requires HTTPS, authentication, origin policy,
46 and usually a TURN service. 47 and usually a TURN service.
47 48
48 Infinite Canvas hosts this page as a hidden CEF/iframe transport. Pressing `M` 49 Infinite Canvas hosts this page as a hidden CEF/iframe transport. Pressing `M`
49 shows partial and final transcript text in a centered Raylib caption; pressing 50 shows partial and final transcript text in the retained Dictation entity;
50 `Enter` sends the accumulated text to Qwen session orchestration: 51 pressing `Enter` sends the accumulated text to Copilot orchestration:
51 52
52 ```bash 53 ```bash
53 bazel run //infinite_canvas:agent_dev 54 bazel run //infinite_canvas:agent_dev
54 ``` 55 ```