chichi-speech
par hudeven
Speech & TranscriptionRésumé
A RESTful service for high-quality text-to-speech using Qwen3.
Description
Ce que fait ce skill
A FastAPI-based REST service that converts text to speech using the Qwen3 TTS model. Designed around voice cloning with a cached reference audio prompt, so the voice stays consistent and generation stays fast across requests.
Pre-computing the voice prompt from the reference audio avoids redundant inference on every request, making repeated synthesis faster than services that process the reference each time.
Quand l'utiliser
- Converting written content to audio with a consistent cloned voice
- Running a local TTS endpoint for integration into other tools or scripts
- Testing voice cloning with custom reference audio files
- Generating speech in multiple languages via a simple HTTP call
- Embedding speech synthesis into an application via REST API
Exemple de workflow
User asks: synthesize 'Nice to meet you' in English
- 1
Start the chichi-speech server on port 9090 with a reference audio URL and reference text
- 2
Confirm the service is healthy by hitting the /docs endpoint
- 3
Send a POST request to /synthesize with the text and language fields
- 4
Receive the synthesized audio as a WAV file
output/nice_to_meet.wav containing the cloned-voice speech
Prérequis
- Reference audio file (URL or local path) for voice cloning if using a custom voice
Installation
clawhub install hudeven/chichi-speechExécute la commande CLI, ou colle l'instruction dans Claude Code, Cursor, Codex ou l'agent de code que tu utilises.