AI · AI model
NVIDIA Parakeet TDT 0.6B V2
Open-source 600M English speech-to-text model with fast inference and timestamps
Parakeet TDT 0.6B V2 is an open automatic speech recognition model from NVIDIA based on FastConformer architecture. It transcribes audio up to 24 minutes in a single pass while outputting punctuation, capitalization, and word-level timestamps. It achieves high benchmark accuracy under a commercially permissive CC-BY-4.0 license.
What it's for
- Transcribe recorded audio and podcasts with word-level timestamps
- Generate accurate subtitles and captions for video content
- Power speech-to-text pipelines in voice bots and conversational AI
- Transcribe phone calls and meeting recordings at high speed
Features
- FastConformer-TDT architecture with 600M parameters
- Word, character, and segment-level timestamp generation
- Automatic punctuation and capitalization in transcriptions
- Single-pass processing for audio files up to 24 minutes
- High throughput with up to 3380 RTFx in batched inference
- Deployable via NVIDIA NeMo or accessible via hosted Riva API
- Noise robustness tested across multiple SNR levels
Pricing
Open source. Free and open source under CC-BY-4.0; hosted API available on NVIDIA Build.
Details
- Type
- AI model
- Category
- AI
- Topic
- Text to Speech
- Open source
- Yes
- Platforms
- Linux, Code library, API, Self-hosted, Command line
- For
- Developers, ML engineers, and researchers building speech-to-text and transcription systems.
- Alternative to
- OpenAI Whisper, ElevenLabs Scribe, Deepgram Nova-2
- Tags
- asr, speech to text, audio, nvidia, nemo, transcription, fastconformer
Why it was saved
- @therundownai: Achieved top ranking on the Open ASR Leaderboard with a 6.05% Word Error Rate (surpassing Whisper and ElevenLabs' Scribe) and is available under a commercially permissive CC-BY-4.0 license.
Related
AI voice generator, voice cloning, and conversational agent platform
Ultra-low-latency streaming text-to-speech and instant voice cloning model.
Open-source 1.6B text-to-speech model generating multi-speaker dialogue in one pass.
Fast, low-cost text-to-speech API and studio for creators and developers
Converts spoken thoughts into structured, polished writing in customizable styles.
Official Python library for Cartesia's real-time text-to-speech API
Updated