POST
/
speech-to-text
curl --request POST \
  --url https://api.sarvam.ai/speech-to-text \
  --header 'Content-Type: multipart/form-data' \
  --form model=saarika:v2 \
  --form language_code=unknown \
  --form with_timestamps=false \
  --form with_diarization=false \
  --form num_speakers=123
{
  "request_id": "<string>",
  "transcript": "नमस्ते, आप कैसे हैं?",
  "timestamps": {
    "timestamps": {
      "end_time_seconds": [
        16.27
      ],
      "start_time_seconds": [
        0
      ],
      "words": [
        "Good afternoon, this is Naveen from Sarvam."
      ]
    }
  },
  "diarized_transcript": {
    "entries": [
      {
        "transcript": "<string>",
        "start_time_seconds": 123,
        "end_time_seconds": 123,
        "speaker_id": "<string>"
      }
    ]
  },
  "language_code": "<string>"
}

Headers

​
api-subscription-key
string
default:

Your unique subscription key for authenticating requests to the Sarvam AI Speech-to-Text API. Here are the steps to get your api key

Body

multipart/form-data
​
file
file
required

The audio file to transcribe. Supported formats are WAV (.wav) and MP3 (.mp3). The API works best with audio files sampled at 16kHz. If the audio contains multiple channels, they will be merged into a single channel.

​
model
enum<string>

Specifies the model to use for speech-to-text conversion. Note:- Default model is saarika:v2

Available options:
saarika:v1,
saarika:v2,
saarika:flash
​
language_code
enum<string>

Specifies the language of the input audio. This parameter is required to ensure accurate transcription. For the saarika:v1 model, this parameter is mandatory. For the saarika:v2 model, it is optional. unknown: Use this when the language is not known; the API will detect it automatically. Note:- that the saarika:v1 model does not support unknown language code.

Available options:
unknown,
hi-IN,
bn-IN,
kn-IN,
ml-IN,
mr-IN,
od-IN,
pa-IN,
ta-IN,
te-IN,
en-IN,
gu-IN
​
with_timestamps
boolean
default:false

Enables timestamps in the response. If set to true, the response will include timestamps in the transcript.

​
with_diarization
boolean
default:false

Enables speaker diarization, which identifies and separates different speakers in the audio. When set to true, the API will provide speaker-specific segments in the response. Note: This parameter is currently in Beta mode.

​
num_speakers
integer | null

Number of speakers to be detected in the audio. This is used when with_diarization is set to true.

Response

200
application/json
Successful Response
​
request_id
string | null
required
​
transcript
string
required

The transcribed text from the provided audio file.

Example:

"नमस्ते, आप कैसे हैं?"

​
timestamps
object | null

Contains timestamps for the transcribed text. This field is included only if with_timestamps is set to true

Example:
{
  "timestamps": {
    "end_time_seconds": [16.27],
    "start_time_seconds": [0],
    "words": [
      "Good afternoon, this is Naveen from Sarvam."
    ]
  }
}
​
diarized_transcript
object | null

Diarized transcript of the provided speech

​
language_code
string | null

This will return the BCP-47 code of language spoken in the input. If multiple languages are detected, this will return language code of most predominant spoken language. If no language is detected, this will be null