GGUF

start your own cantonese tts server

# 1. runtime
pip install ggk

# 2. serve the model
ggk -m cantonese-1.2-0.6b-q8_0.gguf \
  --adapter pig_cantonese_0.6b_adapter-test-f16.gguf \
  --codec-model tts-tokenizer-12hz-q8_0.gguf \
  --serve --port 8888

# 3. speak
curl 127.0.0.1:8888/v1/audio/speech \
  -H "Content-Type: application/json" \
  -d @request.json -o speech.mp3

or copy and paste the prompt below to your agent

Read https://cantonese.dev/llms-full.txt and follow it:
install ggk, start the Cantonese TTS server on port 8888,
then synthesise "今日天氣好好,我哋出去行下街啦。" with a warm
female voice and save it as speech.mp3. Send the request
body as UTF-8.

or try it from the site

https://cantonese.dev

Downloads last month
56
GGUF
Model size
0.9B params
Architecture
pig
Hardware compatibility
Log In to add your hardware

8-bit

16-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for cantonese-dev/tts-server

Quantized
(1)
this model