Gradio

EYEDOL ASR Demo

Select model, upload audio or use the microphone. Supports long audio via chunking or model long-form timestamps.

Choose model

Record (click Record → Stop)

Or upload audio file

Input source

Use microphone input Use uploaded file

Try model's built-in long-form timestamps (if supported)

Chunk length (seconds)

10 120

Chunk overlap (seconds)

0 5

Note: If a model is private add HF_TOKEN as a secret in Space settings. GPU recommended for best performance.

Full transcription

Segments (start_s, end_s, text)