FrontierThe story, in brief
Meet S1-mini: Superwhisper’s 462 MB Open-Weights Text Normalizer That Turns Raw ASR Transcripts Into Clean Written Text
462 MB. That's all Superwhisper needs to clean messy speech-to-text — and it runs locally.
The KeyNews take
Why it matters
Open-weight text normalizer addresses a real post-ASR gap: converting raw transcripts into publishable text without cloud dependency. Practitioners building speech pipelines now have a lightweight, deployable alternative to proprietary normalization.
The key facts
9 to knowS1-mini: 462 MB open-weights model
Post-ASR text normalizer (removes fillers, self-corrections)
Runs locally (no cloud dependency)
Superwhisper release
Published August 2026
Post-ASR text normalization (filler removal, self-correction resolution)
Local inference capability
Released by Superwhisper
Published August 20, 2026
Go to the source
MarkTechPostmarktechpost.com
Publisher excerpt: S1-mini is a 462 MB open-weights normalizer that sits after ASR, removing fillers and resolving self-corrections locally.