Partner test access

For invited evaluation partners. Paste your access token, pick one or more models, and try a sample or your own recording.

Access token Models — pick one, two or all three; they run side by side

Chinar-Fevery call at volume Chinar-Ltranscripts people read Chinar-Pno GPU needed

Decoding (Chinar-L)

Try a sample

Eight short Azerbaijani recordings with their human transcripts, so you can compare at a glance.

Sample recordings and transcripts: LocalDoc Azerbaijani ASR, CC BY 4.0. Converted to MP3.

Your own recording

Audio file (WAV, MP3, M4A, OGG; up to 50 MB)

Audio is held only in memory while it is transcribed. It is never stored and never used for training. API: POST /v1/audio/transcriptions with Authorization: Bearer <token> and form fields file, model, profile.