CoolFace
Apppublic

SFM2001/spititout

sourceHugging Faceupdated 5mo agoView on Hugging Face
0likes
App README

<div align="center"> <img width="1200" height="475" alt="GHBanner" src="https://github.com/user-attachments/assets/0aa67016-6eaf-458a-adb2-6e31a0763ed6" /> </div>

SPITITOUT

This version is set up for a Hugging Face Docker Space. It does not use Gemini or any external model API; the Space loads local Hugging Face models for chat, speech recognition, and speech synthesis.

Models

  • —Text on CPU Space: Qwen/Qwen3-1.7B-GGUF with llama.cpp / GGUF quantization
  • —Text on GPU Space: Qwen/Qwen3-4B-Instruct-2507 with Transformers, or vLLM if you add a dedicated vLLM server
  • —Speech to text: openai/whisper-tiny
  • —Text to speech: hexgrad/Kokoro-82M through the kokoro Python package, Mandarin voice zf_xiaobei

For a CPU-only Space, keep LLM_BACKEND=llamacpp. To switch back to Transformers, set LLM_BACKEND=transformers and update TEXT_MODEL.

Run locally

bash
npm install
npm run build
pip install -r requirements.txt
python app.py

Open http://localhost:7860.

Deploy to Hugging Face Spaces

Create a Docker Space, then push this folder. The included Dockerfile builds the React frontend and serves it from the FastAPI backend.