Running
22
CrisperWhisper 2.0 Controllable ASR
🎙
Verbatim + intended transcripts with word-level timing
AI & ML interests
Speech intelligence
Recent Activity
View all activity
Papers
Transcription Policy as a Latent Variable: Activating Controllable Verbatim ASR with Word-Level Timing
Content is What Remains: Invariant Speech Tokenization from Parallel Utterances
Organization Card
Speech carries more than words. nyra labs, the research lab of nyra health, builds open speech models that preserve what was said and measure when, who, and how - so real human speech can be understood in full.
models 12
nyralabs/nyra_forced_aligner_en
Updated • 1 • 6
nyralabs/nyra_forced_aligner_en_pro
Updated • 3
nyralabs/CrisperWhisper2.0_small
Automatic Speech Recognition • 0.2B • Updated • 4.33k • 9
nyralabs/CrisperWhisper2.0_medium
Automatic Speech Recognition • 0.8B • Updated • 2.83k • 3
nyralabs/CrisperWhisper2.0_turbo
Automatic Speech Recognition • 0.8B • Updated • 4.8k • 13
nyralabs/CrisperWhisper2.0_large
Automatic Speech Recognition • 2B • Updated • 18.5k • 118
nyralabs/CrisperWhisper2.0_small_pro
0.2B • Updated • 2
nyralabs/CrisperWhisper2.0_medium_pro
0.8B • Updated • 2 • 2
nyralabs/CrisperWhisper2.0_turbo_pro
0.8B • Updated • 7
nyralabs/CrisperWhisper2.0_large_pro
2B • Updated • 23