Models
Open models for speech as it is spoken
Start with verbatim recognition. Add timing, speakers, prosody, and interpretable speech representations as each layer of the research program ships.
live · dual transcript · word-timed
or try a sample:
Record yourself, upload audio or pick a sample - both transcripts come back from one pass.
Run it
pip install "crisperwhisper[ct2]"
from crisperwhisper import CrisperWhisperModel
model = CrisperWhisperModel()
result = model.transcribe("meeting.wav", language="en")
print(result.text)Open weightsnyralabs/CrisperWhisper2.0_large
on Hugging Face
Free for research. For products, licensing takes minutes - model license.
Open weights on Hugging Face
Our released models include weights and model cards for research, evaluation, and fine-tuning.
Benchmarks you can rerun
We publish the benchmarks we measure ourselves against, with the data to reproduce every number.
Licensing for products
Research use is free. Commercial use requires one straightforward license - not a sales cycle.
All models
- CrisperWhisper 2.0ShippedVerbatim speech recognition that keeps fillers, repairs, and pauses - with word-level timestamps.
- nyra forced alignerIn developmentPhoneme-level timing, robust to spontaneous, disfluent, real-world speech.
- PINTPublishedOur parallel-invariant tokenizer: the content of speech, with everything else factored out.
- Fully factorized codecsIn progressContent, speaker, prosody, and conditions - each in its own place, at its own granularity.
Put real speech in your product
Run CrisperWhisper 2.0 in your browser, pull the weights from Hugging Face, or talk to us about a commercial license.