curl --fail-with-body https://api.chariot.in/v1/asr/transcribe \
-H "chariotai-api-key: $CHARIOT_API_KEY" \
-F "input_audio_file=@recording.wav"{
"request_id": "a38c6b32-1f3a-4f02-a779-478282bce082",
"created": "2026-10-23T07:00:00Z",
"text": "Hi, I'd like to move the meeting to Thursday.",
"duration": 4.82,
"words": [],
"model_type": "chariot-1-asr-realtime"
}Transcribe a file
Upload a recording or provide an audio URL and receive a complete transcript.
curl --fail-with-body https://api.chariot.in/v1/asr/transcribe \
-H "chariotai-api-key: $CHARIOT_API_KEY" \
-F "input_audio_file=@recording.wav"{
"request_id": "a38c6b32-1f3a-4f02-a779-478282bce082",
"created": "2026-10-23T07:00:00Z",
"text": "Hi, I'd like to move the meeting to Thursday.",
"duration": 4.82,
"words": [],
"model_type": "chariot-1-asr-realtime"
}Authorizations
Your Chariot API key, generated from the dashboard. Send it in the chariotai-api-key header on every request.
Body
- File upload
- Media URL
Send exactly one of input_audio_file or input_audio_url as multipart/form-data. Other fields apply to either input.
The audio or video file bytes. The file must contain a readable audio track. Supply this or input_audio_url, not both.
Requested transcription language. Omit or use none for automatic detection. This does not request translation.
none, en, hi, bn, gu, kn, ml, mr, or, pa, sa, ta, te File transcription always uses high_accuracy. balanced is only supported by the live WebSocket.
high_accuracy Writing style for mixed-language speech. codemix keeps borrowed words in their own script; native writes them in the spoken language's script.
codemix, native Return estimated word timestamps in seconds from the start of the file. Applies to either file upload or URL input.
Public ASR model. The same model_type is used across WebSocket, file and batch transcription.
chariot-1-asr-realtime Response
Complete transcription.
One complete file transcript; uploaded channels are mixed to mono.
Identifier for this transcription request.
Response creation time in UTC.
Complete recognized text, empty when no speech is recognized.
Processed audio duration in seconds, not request latency.
Public streaming ASR model_type values, the same parameter name TTS
uses. A product name, never the checkpoint or serving name behind it: the
upstream's served name is configuration and is never sent to a caller.
chariot-1-asr-realtime Estimated word boundaries in seconds from the start of the file when include_timestamps=true. Null when requested but unavailable or invalid; an empty list means no timed words. Also empty when timestamps are not requested.
Show child attributes
Show child attributes