NVIDIAOct 10, 2025
Parakeet ASR
Parakeet TDT 0.6B v3 is NVIDIA’s open-weight speech recognition model, with 600 million parameters and support for 25 European languages, which it detects on its own. It transcribes quickly, with punctuation and capitalization.
$0.36
per hour of audio
$0.006
per minute
25 MB
per file
One file, up to 25 MBminutes of audio
Opus in OGG or WebM32 kbps109
M4A or AAC64 kbps55
MP3128 kbps27
WAV16 kHz mono14
WAV44.1 kHz stereo2.5
Split a longer recording and send each part.
Quickstart
OpenAI-compatible: send the recording as a multipart file to the Venice base URL with a Venice API key.
curl https://api.venice.ai/api/v1/audio/transcriptions \
-H "Authorization: Bearer $VENICE_API_KEY" \
-F model=nvidia/parakeet-tdt-0.6b-v3 \
-F file=@meeting.mp3Base URL
https://api.venice.ai/api/v1Parameters
| Parameter | Values |
|---|---|
file | wav, flac, m4a, aac, mp4, mp3, ogg, webm |
model | nvidia/parakeet-tdt-0.6b-v3 |
response_format | json | text |
timestamps | true | false |
language | An ISO 639-1 code, such as en |
Capabilities
Formatswav · flac · m4a · aac · mp4 · mp3 · ogg · webm
Timestamps
timestamps: trueLanguage hint
language: "en"OpenAI-compatible
client.audio.transcriptions.create()SubtitlesBuild them from timestamps; there is no SRT or VTT output
Failures are freeA failed transcription is not charged
Pricing
Billed per second of audio, in USD.
Specs
- Model ID
nvidia/parakeet-tdt-0.6b-v3- Endpoint
POST /api/v1/audio/transcriptions- Formats
- wav, flac, m4a, aac, mp4, mp3, ogg, webm
- File
- Up to 25 MB, sent as multipart
- Output
jsonortext- Timestamps
- Word or segment, depending on the model
- Privacy
- Private
- Weights
- nvidia/parakeet-tdt-0.6b-v3
- Added
- Oct 10, 2025
Compare
Every speech-to-text model on Venice, cheapest first.
| Model | Privacy | Per hour |
|---|---|---|
| xAI Speech to Text v1 | Private | $0.1133 |
| Parakeet ASR | Private | $0.36 |
| Whisper Large V3 | Private | $0.36 |
| Wizper (Whisper v3) | Private | $0.36 |
| ElevenLabs Scribe V2 | Anonymized | $0.6012 |