Whisper Timestamped
Whisper extension providing word-level timestamps for transcription.
About
whisper-timestamped extends OpenAI Whisper with per-word timestamps and per-word confidence scores using dynamic time warping over the model's cross-attention weights. It is compatible with any openai-whisper version, runs the same models, and adds heuristics to suppress common hallucinations on near-silent segments. Useful for subtitle generation, karaoke alignment, and audio editing workflows. Released under the AGPL-3.0 license.
Reviews (0)
Leave a Review
No reviews yet. Be the first to review!
Details
- Price
- Free
- Platform
- Local/Desktop
- Difficulty
- Easy (2/5)
- License
- LGPL-3.0
- Minimum VRAM
- 4 GB
- Added
- Apr 3, 2026
Related Tools
Multilingual ASR model by NVIDIA supporting 4 languages with translation.
Convolution-augmented transformer for speech recognition in ESPnet toolkit.
Pre-trained speech models for STT, TTS, and VAD with simple PyTorch integration.
CLI tool that transcribes audio 10x faster using pipeline optimizations.
Self-supervised speech representation model by Meta for ASR.
Open-source speaker diarization and voice activity detection toolkit.