mistral-voxtral #2

Merged

florian merged 2 commits from mistral-voxtral into main

2026-10-05 12:27:51 +02:00

florian commented

2026-10-05 12:27:47 +02:00

Owner

No description provided.

florian added 2 commits

2026-10-05 12:27:47 +02:00

Add Mistral (Voxtral) as its own speech-to-text API c5252f7c43

Mistral's /audio/transcriptions looks OpenAI-compatible but ignores
OpenAI's response_format/diarized_json and timestamp_granularities[]
fields, so transcripts came back as plain text. The Mistral dialect sends
diarize and timestamp_granularities=segment as Mistral documents, and
the parser reads its speaker_id field. Mistral can't combine timestamps
with a fixed language, so the language field is disabled for it.

Speakers are now numbered in order of appearance (Speaker 1, 2, ...)
for every API, since each labels them differently.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Ask before discarding unsaved speech-to-text settings c14534f9c5

Leaving Settings (back gesture or the top bar arrow) with edited
speech-to-text fields now asks to save, discard or keep editing,
instead of silently dropping the edits. Save is only offered when the
form is valid. "Edited" is measured against what the form showed when
it opened, so the pre-filled defaults on a fresh install don't trigger
the prompt.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

florian merged commit 7949986268 into main

2026-10-05 12:27:51 +02:00

florian deleted branch mistral-voxtral

2026-10-05 12:27:51 +02:00

florian referenced this pull request from a commit

2026-10-05 12:27:52 +02:00

Merge pull request 'mistral-voxtral' (#2) from mistral-voxtral into main