glm-4-voice-decoder-emo-ft

Built with glm-4.

Fine-tuned GLM-4-Voice decoder weights for emotion-preserving Chinese ↔ English speech-to-speech translation, used together with the Kimi-Audio Emotion-Aware S2ST training / inference pipeline.

Files

File	Size	Role
`epoch500_emoft.pt`	~425 MB	Fine-tuned flow checkpoint (emotion-preserving)
`hift.pt`	~79 MB	HiFT vocoder checkpoint

Usage

git clone https://github.com/<YOUR_GH_USER>/kimi-audio-release
cd kimi-audio-release
./scripts/download_weights.sh
# the two files will be placed under glm_4_voice_decoder/

'EOF'

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

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