Skip to main content
POST
ElevenLabs Speech to Text V1
Transcribe audio or video files. When use_multi_channel is true and the uploaded audio has multiple channels, a ‘transcripts’ object is returned, with one transcript per channel. Otherwise, a single transcription result is returned.

Request Headers

string
required
Enum value: application/json
string
required
Bearer authentication format: Bearer {{API key}}.

Request Body

integer
If specified, the system will make a best effort to sample deterministically. Requests with the same seed and parameters should return the same result, but absolute determinism is not guaranteed. Must be an integer between 0 and 2147483647.Value range: [0, 2147483647]
boolean
default:false
Whether to annotate the current speaker in the uploaded file.
string
default:"other"
The input audio format. Options are ‘pcm_s16le_16’ or ‘other’. pcm_s16le_16 requires the audio to be 16kHz sample rate, 16-bit integer, mono, little-endian format, and has lower latency compared to encoded waveforms.Allowed values: pcm_s16le_16, other
number
Controls the randomness of the transcription output. The value ranges from 0.0 to 2.0; higher values make results more diverse and less deterministic. If omitted, the selected model’s default temperature is used (usually 0).Value range: [0, 2]
integer
The maximum number of speakers in the uploaded file. This can be used to help distinguish speakers, with support for up to 32 speakers.Value range: [1, 32]
string
Specify the ISO-639-1 or ISO-639-3 language code of the audio file. Providing this in advance can sometimes improve transcription performance. The default is null, which automatically detects the language.
boolean
default:true
Whether to tag audio events such as (laughter) and (footsteps) in the transcription.
string
required
The HTTPS link to the file to be transcribed. Exactly one of file and cloud_storage_url must be provided. The file must be accessible over HTTPS and smaller than 2GB. Any valid HTTPS address is supported, including cloud storage (AWS S3, GCS, Cloudflare R2, etc.), CDNs, or other HTTPS sources. Presigned links with tokens or authentication via URL query parameters are supported.
boolean
default:false
Whether the audio file is multi-channel and each channel contains only a single speaker. When enabled, each channel is transcribed independently and the results are combined. Each word in the output content contains a channel_index field. Up to 5 channels are supported.
number
The diarization threshold. A higher value lowers the probability that one person is split into multiple speakers, but increases the probability that different people are merged into one speaker (fewer speakers detected). A lower value increases the probability that one person is split into multiple speakers, but reduces the probability that different people are merged into one speaker (more speakers detected). Can only be set when diarize=True and num_speakers=None. The default is None, and the threshold is selected based on the model id (usually 0.22).Value range: [0.1, 0.4]
string
default:"word"
The granularity of timestamps in the transcription. ‘word’ provides word-level timestamps, while ‘character’ provides timestamps for each character.Allowed values: none, word, character

Response Information

The response may be one of the following response types:
string
required
The raw transcribed text.
object[]
required
A list of words and their timing information.
integer
The channel index corresponding to this transcript (valid for multi-channel audio).
string
required
The detected language code (for example, ‘eng’ for English).
string
The unique transcription ID for this response.
number
required
The confidence of language detection (between 0 and 1).
object[]
required
A list of transcripts corresponding to each audio channel. Each transcript contains the text for its channel and word-level details.
string
The unique transcription ID for this response.