Speech to transcript to a clean, structured clinical note, most of it running right in your browser. Review it, fix anything, and export to PDF, Word, or Markdown. The patient's audio never has to touch a server.
Fluent in clinical shorthand
Hosted GPUs through Hugging Face. Reach for the heavy hitters, Llama 3.3 70B, GPT-OSS 20B, Gemma, when you want the strongest structured output and you're fine sending text to a provider.
Best quality · a few seconds per note
Whisper, SmolLM, Qwen, and Gemma running in your browser over WebGPU or WASM. The input never leaves the machine. Free every time, and oddly satisfying to watch the tokens land.
Private · nothing leaves the device
Record the consultation or paste a transcript. With live transcription on, the words show up as they are spoken.
Pick any model. NeoScribe turns the back-and-forth into complaints, history, examination, assessment, and plan.
Read it back, correct anything, and download a tidy note as PDF, Word, or Markdown to drop into the record.
Whisper and small language models run in the browser over WebGPU or WASM. The recording and the transcript stay on the machine unless you reach for a cloud model.
Run the same consult through two or three models at once and see who caught the diagnosis, the dose, and the follow-up.
When a model is too heavy for your device or a provider is having a bad day, NeoScribe tells you in plain English instead of spinning forever.
Transcripts and notes are saved and searchable, so last Tuesday's clinic is one click away when the paperwork catches up with you.
The default path keeps the recording and the transcript on your device. Whisper does the listening locally, and you decide if anything ever leaves. When you want a hosted model, that's one click, and NeoScribe is upfront about what gets sent.
Open NeoScribe