Skip to content

Changelog

Changelog for the Amazon Linux 2023-based WhisperX images.

WhisperX 3.8.6 — 2026-08-14

Tags: 3.8.6-cu128-amzn2023 · 3.8.6-cu128-amzn2023-sagemaker

WhisperX source: v3.8.6

Highlights

  • Initial release of WhisperX containers on Amazon Linux 2023.
  • Speech transcription, word-level alignment (wav2vec2), and speaker diarization (pyannote) through an OpenAI-compatible API.
  • Deployable on EC2 (port 8000) and Amazon SageMaker AI (port 8080, real-time and asynchronous endpoints).
  • Built on CUDA 12.8 with Python 3.12; faster-whisper 1.2.1, CTranslate2 4.8.0, and pyannote.audio 4.0.7.

Supported Models at Launch

  • Whisper tiny, base, small, medium, large-v2 (default), and large-v3 — served one per container via WHISPERX_DEFAULT_MODEL.
  • Per-language wav2vec2 alignment models, downloaded on demand.
  • pyannote speaker-diarization-community-1 diarization pipeline, baked into the image.