Modelpublic

omi-health/omi-med-stt-v1

sourceHugging Facecc-by-4.0updated 24d agoView on Hugging Face

7likes2.7kdownloads

Model Card

Omi Med STT v1

Omi Med STT v1 is an English medical speech-to-text model for clinical dialogue, built from NVIDIA Parakeet TDT 0.6B v2.

It is designed for local transcription of GP-style consultations, medication review, clinical dictation, and procedure/device/test language. The canonical checkpoint is this NeMo .nemo model. Mac and CPU exports are available in the companion repos below.

Why Use It

—Open-weight, 0.6B-class medical ASR model
—Runs locally through the omi-med-stt CLI
—Stronger medical transcription than base Parakeet v2 in Omi's internal evaluation
—Runtime options for Apple Silicon, NVIDIA CUDA, and Linux/Windows CPU

Quickstart

Recommended CLI:

bash

pip install -U omi-med-stt
omi-med-stt audio.wav

Apple Silicon:

bash

pip install -U "omi-med-stt[mlx]"
omi-med-stt audio.wav

NVIDIA CUDA / canonical NeMo checkpoint:

bash

pip install -U "omi-med-stt[nemo]"
omi-med-stt audio.wav --runtime nemo

Linux or Windows CPU:

bash

pip install -U omi-med-stt
omi-med-stt install-cpp --cpp-backend cpu
omi-med-stt audio.wav --runtime cpp

Direct NeMo use:

python

from huggingface_hub import hf_hub_download
import nemo.collections.asr as nemo_asr

nemo_path = hf_hub_download("omi-health/omi-med-stt-v1", filename="omimedstt-v1.nemo")
model = nemo_asr.models.ASRModel.restore_from(nemo_path)
print(model.transcribe(["audio_16k_mono.wav"])[0])

Which Repo Should I Use?

Use case	Repo	Notes
Best quality / NVIDIA GPU	`omi-health/omi-med-stt-v1`	Canonical NeMo checkpoint
Apple Silicon default	`omi-health/omi-med-stt-v1-mlx-q8`	Smaller MLX q8 export
Apple Silicon full precision	`omi-health/omi-med-stt-v1-mlx`	Larger MLX export
Linux/Windows CPU	`omi-health/omi-med-stt-v1-gguf`	GGUF q8_0 via parakeet.cpp

Evaluation

The public weights are unchanged. These September 2026 results come from the current documented runtime recipes on the same frozen 1,513-clip, 7.18-hour medical benchmark and scorer. No dictionary, custom vocabulary, contextual bias, reference-aware selection, or transcript correction was used.

Runtime artifact	Platform	WER	M-WER	Drug M-WER	Medical recall
NeMo canonical	NVIDIA CUDA (L4, BF16)	6.54%	2.23%	4.75%	97.77%
MLX q8	Apple Silicon (M4 Max)	6.65%	2.12%	4.52%	97.88%
GGUF q8_0	Linux/Windows CPU	7.10%	2.16%	4.30%	97.84%

Lower is better for WER, M-WER, and Drug M-WER; higher is better for medical recall. GPU is the quality and throughput path, MLX q8 is the selected Apple runtime, and CPU is the portable fallback. The CPU row has the lowest observed occurrence counts in this draw, but paired tests do not establish it as clinically better than GPU or MLX.

Compared with the open-model rows on Omi's standing 30-system board, the CUDA and MLX q8 runtimes have the lowest observed WER, while MLX q8 has the second-lowest observed M-WER. These are positions in this benchmark draw, not a universal ranking.

The previous model-card figures were produced by older runtime/scorer paths. The numbers above supersede them as runtime results; they do not represent a new training run or changed model weights.

Notes

—The model is English-only.
—The CLI converts audio to 16 kHz mono automatically.
—The CPU GGUF path is for portability; use MLX or NeMo when available.

Safety

Omi Med STT v1 is speech-to-text only. It is not a diagnostic, triage, prescribing, or clinical decision model, and it is not clinically validated. Transcripts must be reviewed before any clinical use.

License And Attribution

Weights are released under CC-BY-4.0, matching the base model.

Omi Med STT v1 is a derivative of `nvidia/parakeet-tdt-0.6b-v2`. It is not an NVIDIA model.

Runtime: Omi-Health/omi-med-stt-runtime under MIT.

Built by Omi Health. Related product: Omi Scribe.