OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Audio & Voice AI

Whisperx

WhisperX is an open-source Python speech-recognition library with a command-line interface for transcribing audio, aligning words to timestamps and assigning speaker ID labels to speech segments.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Whisperx — OpenViST project profile in Audio & Voice AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use WhisperX when an audio-transcription workflow needs word-level timing or speaker-turn labels, rather than only utterance-level timestamps.

Who it is for

  • People processing recorded audio who need word-level transcript timing.
  • Developers building transcription or audio-processing workflows.
  • Users who need speaker ID labels in transcripts and can provide the required Hugging Face token and model-agreement acceptance.

Documented uses

  • Transcribe an audio file from the command line.
  • Align recognized words to more precise timestamps.
  • Create transcripts with segments labeled by speaker ID.
  • Highlight word timings in an SRT file.
Before you choose

Requirements & limits

  • GPU acceleration requires CUDA Toolkit 12.8; CPU-only use can skip CUDA.
  • Diarization requires a Hugging Face read token and acceptance of the speaker-diarization-community-1 model agreement.
  • The README notes that larger models may improve timestamp accuracy but require more GPU memory.
  • The README reports a 70x real-time result with large-v2; actual performance will depend on the system and workload.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install WhisperX and transcribe an audio file

Install the CLI and run the documented transcription command on an audio file.

Before you start

  • Python package installation using pip, or uvx to run the tool.
  • For GPU acceleration, CUDA Toolkit 12.8 installed first; skip CUDA for CPU-only use.
  • The README notes ffmpeg and Rust may also be needed.

Steps

  1. Install with the command below, or use uvx whisperx.

    Documented command
    pip install whisperx

  2. Run whisperx path/to/audio.wav with the path to the audio file.

Expected outcome

WhisperX processes the audio for transcription; word-level timing is produced through its alignment workflow.

Guide limitations

  • The README says larger models can improve timestamp accuracy while using more GPU memory.
  • Additional dependencies such as ffmpeg and Rust may be needed.
Documentation 1 ↗
Source-based practical guide

Prepare diarization for a transcription

Complete the documented access and model-agreement prerequisites for speaker diarization.

Before you start

  • A Hugging Face read-access token.
  • Acceptance of the speaker-diarization-community-1 model agreement.

Steps

  1. Generate a read-access token through Hugging Face.

  2. Accept the user agreement for the speaker-diarization-community-1 model.

  3. Pass the token after the --hf_token argument when enabling diarization.

Expected outcome

The documented diarization prerequisites are in place to label transcript segments with speaker IDs.

Guide limitations

  • The README describes speaker ID labels, not identification of speakers by their real names.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
14 October 2026
Original website title
m-bain/whisperX · Official homepage title · retained in its source language
How we review sourcesSuggest a correction