Free🇨🇳
Open-source speech recognition toolkit from Alibaba (Tongyi Lab, ModelScope) with streaming, speaker diarization and emotion recognition.
WhisperX overview and key facts (October 2026)
Open-source speech recognition based on Whisper with word-level timestamps and speaker diarization.
Free
What is WhisperX?
WhisperX is an open-source speech recognition tool that extends OpenAI's Whisper with fast batched transcription, precise word-level timestamps and speaker labels, for teams that transcribe recordings on their own hardware.
WhisperX uses the faster-whisper back end and voice activity detection to split audio and batch it, which the README says reaches 70 times real time with Whisper large-v2 on less than 8 GB of GPU memory. Forced alignment with wav2vec2 adds per-word timing, and pyannote-audio labels speakers. Alignment models exist for languages such as English, German, French, Spanish and Italian. It installs with pip and runs on GPU or CPU. The GitHub API showed about 24,000 stars and a push on September 26, 2026.
WhisperX AI features and use cases
- Transcribes long recordings quicklyBatched inference and voice activity detection process hours of audio in minutes on a single GPU, according to the README. Teams can transcribe interviews or meetings in-house instead of paying per minute.
- Marks who said whatSpeaker diarization labels each segment with a speaker. Minutes and interview notes become easier to review and quote.
- Aligns text with exact timingWord-level timestamps make it possible to jump to a word in the recording or create precise subtitles. Editors and trainers save time when cutting video or checking quotes.
What sets WhisperX apart
- Combines Whisper transcription, wav2vec2 alignment and pyannote diarization in one pipeline.
- Runs fully on your own hardware; speaker models need a Hugging Face token but no cloud upload of audio.
Where WhisperX is strongest
Available internationally
Open-source software with no country restriction that runs on your own infrastructure anywhere.
Questions to ask before you buy WhisperX
Recordings of people are personal data, so clarify consent and retention before transcribing meetings or interviews. Speaker diarization models require accepting their terms on Hugging Face. The project is maintained by a small team without commercial support, so plan internal ownership.
WhisperX reviews
No reviews yet. Be the first to share how your team uses it.
Every review shows its verification level. We read each one before it goes live. How we check reviews
Questions about WhisperX
Does WhisperX need a GPU?
A GPU with CUDA speeds it up considerably, but the README notes that CPU-only use is possible.
Which languages does WhisperX support?
Transcription covers Whisper's languages; word alignment models are available for languages such as English, German, French, Spanish and Italian.
Is WhisperX free for commercial use?
The code is under the BSD-2-Clause license. Check the terms of the Whisper and pyannote models you use.
WhisperX update historyno updates yet
No changes recorded for WhisperX yet.
Changes the vendor made and confirmed user reports, each checked against the vendor's own pages. We review every tool monthly.
Alternatives to WhisperX
Tools in the same category (Image, video & audio) with the most overlapping markets.
Free🇫🇷
Open-source Python toolkit for speaker diarization (who speaks when), maintained by the French company pyannoteAI.
Free🇺🇸
Open speech recognition model from OpenAI under the MIT license for multilingual transcription, translation to English and language ID.