Meet S1-mini: Superwhisper’s 462 MB Open-Weights Text Normalizer That Turns Raw ASR Transcripts Into Clean Written Text

Superwhisper has released the S1 family of models : S1-Voice, S1-Language, and S1-mini. S1-Voice is a cloud speech-to-text model, and S1-Language is a cloud instruction-following model for cleanup and formatting. The one that is quite interesting outside the app is S1-mini , released with open weights on Hugging Face. S1-mini is a 0.6B text normalizer, not a transcriber and not a chat model. It sits after automatic speech recognition and rewrites raw transcripts as clean written text: fillers removed, self-corrections resolved to what the speaker landed on, punctuation and capitalization applied, and spoken numbers, dates, currency, and email addresses rendered in written form. It is fine-tuned from Qwen/Qwen3-0.6B , covers English only in release v1, and is steered entirely by a three-axis control line placed above the transcript. Superwhisper reports 94.8% token accuracy on a held-out set of 7,519 cases, measured greedy on the quantized build.

Yes, but only S1-mini. S1-mini is published on Hugging Face under Apache 2.0 plus a naming clause. S1-Voice and S1-Language are Superwhisper-hosted services, so they are consumable, not self-hostable.

S1-mini is a text normalizer, not a transcriber and not a chat model. It sits after automatic speech recognition:

audio → ASR (Whisper, Parakeet, …) → S1-mini → clean text

It removes filler words, resolves false starts and self-corrections to the value the speaker landed on, applies punctuation and capitalization, and renders spoken numbers, dates, times, currency, and email addresses in written form. Say “support at superwhisper dot com” and you get support@superwhisper.com.

The model is fine-tuned from Qwen/Qwen3-0.6B . It has 596M unique parameters (0.44B non-embedding), 28 layers, 16 query heads and 8 key/value heads with GQA, and BF16 weights. The Hub sidebar reports 0.8B because the tied embedding is stored twice; the card explains the discrepancy explicitly. Release v1 covers English only, and recommended input is roughly 1,000 tokens.

S1-mini takes a fixed system prompt, then a control line, then the raw transcript:

Styling takes casual , semi-casual , semi-formal , or formal . Structure takes prose or lists . Context takes general or email . All three axes are independent, and every combination was trained. Send values outside those sets, or reword the system prompt, and output can degrade or garble. Note the small mismatch worth knowing: the Superwhisper app exposes a five-stop tone slider that adds a “balanced” preset, while the open weights document four trained Styling values.