Bug 515829

Summary: Feature: Integrate Voxtral (Mistral AI) for Speech-to-text
Product: [Applications] kdenlive Reporter: reportthebug <chris.seiferth>
Component: Title Clips & SubtitlesAssignee: Jean-Baptiste Mardelle <jb>
Status: REPORTED ---    
Severity: wishlist CC: luzpaz
Priority: NOR    
Version First Reported In: unspecified   
Target Milestone: ---   
Platform: Other   
OS: Linux   
Latest Commit: Version Fixed/Implemented In:
Sentry Crash Report:

Description reportthebug 2026-02-10 17:24:16 UTC 
Support Mistral AI’s new open-weight Voxtral models as an alternative STT engine for automatic subtitling.

Key Benefits:
* Higher Accuracy: Outperforms Whisper Large-v3 in speed and Word Error Rate (WER).
* Native Diarization: Built-in speaker identification to automatically label different voices in transcripts.
* Efficiency: Optimized for local hardware; the Mini-3B model provides high-quality results with low VRAM usage.
* Privacy/License: Apache 2.0 license, allowing for fully offline, private processing.

Proposed Integration:
Add "Voxtral" to the STT engine list in Settings > Speech to Text, with model selection (Mini/Small) and a toggle for speaker diarization.
Comment 1 reportthebug 2026-02-10 17:48:45 UTC 
a demo you can find here: https://huggingface.co/spaces/pandora-s/Voxtral-Subtitles