Creative AI Model or AI service
NVIDIA Parakeet 0.6B v2 transcribes English speech with timestamps
NVIDIA’s Parakeet TDT 0.6B v2 is a downloadable English speech recognition model. It can add punctuation and capitalization and return timestamps for parts of a transcript. [1][2]
- License
- Creative Commons Attribution 4.0 for the documented model. [2]
What the model does
NVIDIA described this 600-million-parameter checkpoint in a June 2025 engineering post. Its model card documents English transcription and links the downloadable model. [1][2]
Why it matters
Timestamps let a reviewer go to the source recording to check an uncertain word. They also support interfaces that let readers move between a transcript and audio. [2]
Who can use it
Teams with English interviews, talks or other recordings can test the exact v2 checkpoint. The model card describes it separately from other Parakeet models. [1][2]
Key dates
- NVIDIA profiles this checkpoint in engineering documentation [1]
Ways to use it
When preparing a caption track, compare segment times against the recording and review uncommon words before export. Test the kinds of accents and background noise the tool will encounter. [2]
Model limitations
The publisher says accuracy varies with accent, noise and audio context. This specific version targets English, and the publisher’s speed figures depend on recording length and batch size. [2]
Sources and verification
Catalog evidence checked How we verify records
- Primary source NVIDIA engineering article documenting Parakeet TDT 0.6B v2
- Primary source nvidia/parakeet-tdt-0.6b-v2
Related profiles
Creative AI
Seedance 2.5 (4K API and Flash)
ByteDance (Volcano Engine / BytePlus)
These tools generate video through developer services. The cited sources describe different variants and routes of access.
Creative AI
LongCat-Video-Avatar 1.6
Meituan (LongCat team)
It generates video of an avatar from supported inputs in the documented project.
Creative AI
MiniMax H3-Image
MiniMax
It generates images from prompts through the model architecture described by MiniMax.