// OminiX ASR (speech-to-text), preset-voice TTS with emotion/speed control, and model management via Qwen3 models on Apple Silicon. For voice cloning and custom voice profiles, use mofa-fm. Triggers: voice, transcribe audio, text to speech, speak this, read aloud, model management, download model, 语音识别, 语音合成, 模型管理.
Harnessed audio-artifact starter. Synthesizes a minimal WAV file under audio/ and relies on the workspace contract to deliver it.
Harnessed coding-assistant starter. Produces a unified-diff artifact and a file-list preview under patches/.
Minimal harnessed single-artifact starter. Use as a template for a custom app that produces one deliverable.
Harnessed report-generator starter. Writes a markdown artifact under reports/ and relies on the workspace contract to deliver it.
Recursively crawl websites using headless Chrome. Triggers: crawl, scrape website, 爬取, crawl site, deep crawl, website content.
Deep multi-round web research with parallel fetching. Triggers: deep search, research, 深度搜索, 调研, investigate, deep research.
| name | voice |
| description | OminiX ASR (speech-to-text), preset-voice TTS with emotion/speed control, and model management via Qwen3 models on Apple Silicon. For voice cloning and custom voice profiles, use mofa-fm. Triggers: voice, transcribe audio, text to speech, speak this, read aloud, model management, download model, 语音识别, 语音合成, 模型管理. |
On-device speech-to-text and preset-voice text-to-speech with emotion control plus model lifecycle management, backed by Qwen3 ASR/TTS models running on the local ominix-api server (Apple Silicon).
fm_tts, fm_voice_save, fm_voice_list, fm_voice_delete). This skill
is preset-voice only.voice_synthesize falls through to the macOS built-in say
command, which auto-picks a system voice from the text language and ignores
the prompt parameter.| Tool | Purpose |
|---|---|
voice_transcribe | ASR — WAV/OGG/MP3/FLAC/M4A → text |
voice_synthesize | Preset-voice TTS with optional emotion + speed |
list_models | List loaded + catalog models on the local ominix-api |
download_model | Pull a catalog model to local disk |
load_model | Load a downloaded model into GPU memory |
unload_model | Free a loaded model from GPU memory |
{"audio_path": "voice.ogg", "language": "Chinese"}
{"text": "Hello world", "language": "english", "speaker": "ryan"}
{"text": "我太开心了！", "speaker": "vivian", "prompt": "用兴奋激动的语气说话，充满热情和活力"}
After voice_synthesize returns a file path, deliver the audio with
send_file.
docs/emotion-prompts.mddocs/api-reference.mdvoice_synthesize with a prompt while ominix-api is down — the
fallback (say) silently drops the emotion. Use list_models to confirm
Qwen3-TTS is loaded before relying on emotion control.download_model + load_model after a fresh install — the
catalog model is not loaded until you load it explicitly.