Explore the feeling
in a voice.
A demo for MERaLiON-SER-v1, designed to turn a short speech recording into emotion scores and dimensions of vocal expression.
Record or upload
The prepared app accepts 1–30 seconds of speech, mixes it to mono, and resamples it for the model.
Explore emotion scores
Neutral, happy, sad, angry, fearful, disgusted, and surprised — using the checkpoint’s own label order.
Read vocal expression
Valence, arousal, and dominance on a 0–1 scale, with downloadable JSON results.
Hugging Face declined both ZeroGPU and basic CPU hosting for this account. This free static page preserves the project while compute access is arranged. It does not accept audio or display simulated model results.
Once eligible compute is available, the owner can restore README-gradio.md as README.md, select ZeroGPU, and test the live demo. Python syntax has passed; live inference remains unverified.