A two-stage replay-based post-training method corrects ASR timestamp drift across non-speech gaps while preserving recognition far better than ordinary timestamp fine-tuning.
CrisperWhisper: Accurate Timestamps on Verbatim Speech Transcriptions
2 Pith papers cite this work, alongside 2 external citations. Polarity classification is still indexing.
2
Pith papers citing it
2
external citations · Pith
abstract
We demonstrate that carefully adjusting the tokenizer of the Whisper speech recognition model significantly improves the precision of word-level timestamps when applying dynamic time warping to the decoder's cross-attention scores. We fine-tune the model to produce more verbatim speech transcriptions and employ several techniques to increase robustness against multiple speakers and background noise. These adjustments achieve state-of-the-art performance on benchmarks for verbatim speech transcription, word segmentation, and the timed detection of filler events, and can further mitigate transcription hallucinations. The code is available open https://github.com/nyrahealth/CrisperWhisper.
fields
cs.CL 2years
2026 2representative citing papers
A new framework shows concept subspaces are not unique, estimator choice affects containment and disentanglement, LEACE works well but generalizes poorly, and HuBERT encodes phone info as contained and disentangled from speaker info while speaker info resists compact containment.
citing papers explorer
-
REDDIT: Correcting Model-Generated Timestamp Drift in ASR without Forgetting via Replay-Based Distribution Editing
A two-stage replay-based post-training method corrects ASR timestamp drift across non-speech gaps while preserving recognition far better than ordinary timestamp fine-tuning.
-
A framework for analyzing concept representations in neural models
A new framework shows concept subspaces are not unique, estimator choice affects containment and disentanglement, LEACE works well but generalizes poorly, and HuBERT encodes phone info as contained and disentangled from speaker info while speaker info resists compact containment.