Create transcript from one uploaded audio file.

Transcribe one audio clip (up to the deployment’s length limit, 120 s by default) and return the transcript in a single JSON response, with optional per-word timestamps and speaker attribution. Send the audio as a multipart file upload (file). The language is set per request (language, ja by default).

Options: with_timestamps adds words with start and end times; diarization attributes words to speakers (speaker_id: speaker_1, speaker_2, …) on deployments whose model carries speaker tokens, with word times null unless with_timestamps is also set; kana writes personal names in katakana.

The same route also accepts a JSON body (Content-Type: application/json) with audio_url (an https URL on fal’s CDN or a base64 data: URI), language, and the optional with_timestamps / diarization flags; this is what the fal Playground and queue send. The response and the billed units are the same for both bodies.

In the Python SDK this route is client.asr.transcribe(path, language=..., with_timestamps=..., diarization=...).

Request

This endpoint expects a multipart form containing a file.

filefileRequired

Audio file to transcribe.

file_formatenumOptionalDefaults to other

Input audio format. Use pcm_s16le_16 for raw 16-bit little-endian PCM at 16kHz mono; use other for encoded audio/video files.

Allowed values:

languagestring or nullOptional

Input language. Same format as input-streaming TTS open messages; defaults to the configured server language.

keywordslist of strings or nullOptional

Keywords to use for the transcription. List of strings. Defaults to none.

kanabooleanOptionalDefaults to false

人名をカタカナで書き起こすかどうか。

with_timestampsbooleanOptionalDefaults to false

Include calibrated timestamp items in the response.

diarizationbooleanOptionalDefaults to false

Attribute words and spacing to speakers. Times are null unless with_timestamps is enabled.

Response

Successful Response

language_codestring

language_probabilitydouble

textstring

wordslist of objects

audio_duration_secsdouble or nullOptional

Errors

422

Unprocessable Entity Error