AGENT SKILL
ambient-faster-whisper-stt
Local faster-whisper STT provider for Ambient speech input validation.
- License 许可
- MIT
- 最近更新
- 2026-08-22
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- ambient-faster-whisper-stt
- description
- Local faster-whisper STT provider for Ambient speech input validation.
02 资源文件 4
- 参考文档ambient-cli.json1.7 KB
- 脚本scripts/faster_whisper_transcribe.py2.0 KB
- 脚本scripts/run.mjs7.5 KB
- 脚本tests/smoke.test.mjs4.5 KB
03 SKILL.md 正文
Use this package only through Ambient's STT provider path. It implements:
faster_whisper_transcribe --audio <wav> --language <language> --output-json <path>
Operational notes:
- The default path is intentionally small:
faster-whisper==1.1.1,tiny.en, CPU, int8, beam size 1. - This package is adapter-only. It bundles Ambient descriptor metadata, these instructions, the Node wrapper, the Python runner, and tests; it does not bundle
uv, Python, faster-whisper wheels, CTranslate2 binaries, or model weights. requestsis installed explicitly in the uv environment because the live macOS smoke found the direct faster-whisper import path needed it.- First real transcription may populate local uv/Python/faster-whisper/model caches. Treat health output as runtime-readiness guidance, not proof that all model assets are already downloaded.
- When validating first-run behavior, use
pnpm run stt:faster-whisper-clean-cachefrom the repo root so uv, Python, Hugging Face, and XDG caches are isolated in the validation artifact. - This is setup/control evidence for the Ambient STT product path. Do not present
tiny.enas multilingual quality evidence. - Use a larger model only after setting
AMBIENT_FASTER_WHISPER_MODELintentionally and recording the runtime/quality tradeoff.
04 同仓库更多 Skills
ambient-tinystyler
Create reusable TinyStyler style profiles from user text examples and apply TinyStyler-style text transfer through Ambient CLI.
★ 0
ambient-qwen3-asr
Local Qwen3-ASR speech-to-text provider for Ambient push-to-talk.
★ 0
ambient-minicpm-v-vision
Local MiniCPM-V visual-understanding provider for Ambient screenshot, UI, game-frame, and image evidence.
★ 0
ambient-macos-voice
暂无简介
★ 0
ambient-imagegen
Generate raster images through hosted APIs such as OpenAI, Google Nano Banana Pro, fal/FLUX, Replicate, Stability AI, and Ideogram.
★ 0
hyperframes
Create deterministic authored-motion MP4/WebM/MOV videos from HTML/CSS/JS compositions through Ambient's bundled HyperFrames adapter.
★ 0