Speech Rec
checking ASR…

Drop audio here

or (batch ok)

Who does the transcription. On-device (Whisper/SenseVoice) keeps audio on the machine but needs the local install. Cloud (Deepgram/Groq) needs an API key and sends audio to that provider.

Bigger models are more accurate but slower and larger to download. Only used by the on-device Whisper engine; cloud engines pick their own model. .en models are English-only.

On-device models

To transcribe fully on the machine — audio never leaves — install Python + faster-whisper on the server, then download a model here. This hosted demo has no on-device engine, so the list is empty; use a cloud engine (Deepgram/Groq) instead. On a local or on-prem install, models appear here to download once.

    QA rubric

    Transcribe an audio file to see the result here.