# linto-ai/whisper-timestamped

Multilingual Automatic Speech Recognition with word-level timestamps and confidence

Repository: https://github.com/linto-ai/whisper-timestamped
Canonical: https://ross.abutalabs.com/products/whisper-timestamped
Language: Python
License: AGPL-3.0
License Family: copyleft
Topics: deep-learning, speech, speech-recognition, speech-to-text, asr, machine-learning, python, python3, pytorch, attention-is-all-you-need, attention-mechanism, attention-model, attention-network, attention-seq2seq, attention-visualization, multilingual-models, speaker-diarization, speech-processing, transformers, whisper
Last push: 2026-08-17T06:41:29+00:00

## Health v2 (maintenance only)
Score: 79/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 98, release rhythm 47, longevity 94
- inputs: {"age_days": 1328, "days_push": 16, "days_rel": 358, "gap_med": 12.0, "n_releases_24m": 5}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2841, forks 211 (observed 2026-08-28T04:07:24.744951+00:00)

## What it is
A Python library extending OpenAI's Whisper models to produce accurate word-level timestamps and confidence scores during multilingual speech recognition. It uses Dynamic Time Warping on cross-attention weights with no extra inference steps and low memory overhead.

## Use cases
- transcribe audio with word-level timestamps
- get confidence scores for each transcribed word
- run speech-to-text on long audio files with low memory
- detect spoken language with confidence
- apply voice activity detection before transcription
- generate subtitles with precise word timing

## When to choose
- you need word-level timestamps rather than Whisper's approximate segment timestamps
- you want confidence scores per word or segment
- you want to keep using openai-whisper with minimal changes
- you process long audio files and care about memory usage

## When to avoid
- you need speaker diarization as a first-class feature
- you require a non-Python or non-PyTorch environment
- you need a permissively licensed dependency (it is AGPL-3.0)
- you only need coarse segment-level transcription

## Facets
- artifact type: library
- maturity: active
- function: speech-recognition, machine-learning, deep-learning
- domain: speech-processing, machine-learning
- platform: python, cli, cross-platform
- tags: whisper, word-timestamps, asr, speech-to-text, pytorch, voice-activity-detection, multilingual, confidence-scores, transcription, natural-language-processing

## Member repositories
- linto-ai/whisper-timestamped (main) score 79

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:24.744951+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:37:15.318871+00:00, confidence not recorded.
  - readme: https://github.com/linto-ai/whisper-timestamped (fetched 2026-08-28T04:07:24.744951+00:00, sha d4ec6a70c24b)
  - registry_pypi: https://pypi.org/pypi/whisper-timestamped/json (fetched 2026-08-29T09:52:48.574257+00:00, sha b992e133c42d)
- Data as of 2026-08-30T08:39:29.467469+00:00.