WhisperX

What is WhisperX?

WhisperX is an open-source Python library and command-line tool that extends OpenAI's Whisper for long-form transcription with accurate timing. It transcribes with a faster-whisper backend in batches, which the README puts at about 70 times real time with the large-v2 model, then runs forced phoneme alignment with a wav2vec2-style model so every word gets its own timestamp. Voice activity detection splits the audio before batching, and optional speaker diarization based on pyannote labels who is speaking. You install it with pip or uvx, use CUDA 12.8 for GPU acceleration or run on CPU with int8, and call it from the terminal or from Python in three steps: transcribe, align, assign speakers. The work was published at INTERSPEECH 2023. WhisperX is a self-run tool rather than a hosted service, and the repository lists no price.

Use cases

• Generate subtitles with word-level timing
• Label who spoke in a recording
• Transcribe long audio quickly on a GPU
• Build searchable transcripts with timestamps
• Prepare speech datasets for research
• Transcribe non-English audio with alignment models

Who it's for

• Developers
• Speech researchers
• Podcast and media teams
• Data scientists

Why use WhisperX

• Word-level timestamps are tighter than plain Whisper output, which helps with subtitles and editing.
• Batched inference makes long recordings finish far faster than real time on a capable GPU.
• Diarization and alignment run in the same pipeline, so one script yields timed, speaker-labeled text.
• It is open source and self-run, so audio stays on your hardware.
• Memory can be trimmed with a smaller batch size, a smaller model or int8 compute.

WhisperX features

• Batched inference with a faster-whisper backend
• Word-level timestamps through forced phoneme alignment
• Speaker diarization using pyannote models
• Voice-activity-based segmentation
• Context-aware interleaved batching option
• Optional Numba acceleration for alignment
• Command-line tool and Python API
• CPU mode with int8 compute type
• Language-specific alignment models for many languages
• Highlighted word timing in SRT output

WhisperX integrations

• faster-whisper
• pyannote.audio
• Silero VAD
• Hugging Face
• CTranslate2

How to get started with WhisperX

1. Install WhisperX with pip install whisperx, adding CUDA toolkit 12.8 first if you want GPU acceleration.
2. Run whisperx on an audio file, or load a model in Python with whisperx.load_model.
3. Transcribe in batches, then align the segments with a language-specific alignment model.
4. Pass a Hugging Face token and the --diarize flag to assign speaker labels to words.
5. Write out subtitle or text files, using --highlight_words to visualize word timing in SRT.

WhisperX limitations

• Words without characters in the alignment dictionary, such as "2014." or "£13.60", get no timing
• Overlapping speech is not handled well by Whisper or WhisperX
• The README says diarization is far from perfect
• A language-specific wav2vec2 alignment model is required
• Diarization needs a Hugging Face token and accepted model terms

WhisperX pricing

WhisperX is open source, so you can use and self-host it for free.

Prices can change and may vary by country or exclude tax. Check the official pricing page.

WhisperX alternatives

The main WhisperX alternatives listed on Debuly are whisper-ctranslate2, whisper.cpp, Deepgram, Hume AI, CAMB.AI and Cartesia. The table compares pricing and platforms; order is A–Z.

1. 

   CAMB.AI

   AI dubbing, text-to-speech, translation and voice cloning across 140+ languages, with a studio app and developer API.

   Audio and voiceVideoDeveloper platforms
   FreemiumVisit CAMB.AI (official site)
2. 

   Cartesia

   Voice AI platform with Sonic text-to-speech, Ink speech-to-text and managed voice agents, priced by credits and minutes.

   Audio and voiceDeveloper platformsAutomation and agents
   FreemiumVisit Cartesia (official site)
3. 

   Deepgram

   Voice AI platform with speech-to-text, text-to-speech and Voice Agent APIs, available as a managed API, in-VPC or self-hosted.

   Audio and voiceDeveloper platformsAutomation and agents
   FreemiumVisit Deepgram (official site)
4. 

   Hume AI

   Expressive text-to-speech and empathic voice agent APIs that let developers direct emotion, tone and pacing in generated speech.

   Audio and voiceDeveloper platformsAutomation and agents
   FreemiumVisit Hume AI (official site)
5. 

   whisper-ctranslate2

   Command-line Whisper client on CTranslate2 and faster-whisper, compatible with OpenAI's CLI, with diarization and live transcription.

   Audio and voiceDeveloper platforms
   Open sourceVisit whisper-ctranslate2 (official site)
6. 

   whisper.cpp

   Dependency-free C/C++ port of OpenAI's Whisper speech recognition that runs offline on CPUs, Apple Silicon and many GPUs.

   Audio and voiceDeveloper platforms
   Open sourceVisit whisper.cpp (official site)

WhisperX FAQ

Sources & verification

Facts on this page come from the official sources below, each with the date we read it. Written by Debuly; the vendor did not review this page.

Image credits

• Logo: none collected yet, so a monogram is shown.
• Cover: none collected yet, so a generated tile is shown.

Spot an error? Report an issue with this listing. We also remove a logo or image on the owner's request.