ONNX Model Zoo

burn874/transcribe-NPU

By burn874 · ↓ 0 · ♥ 0 · Updated 2025-07-22

What's in this repo

This page lists the files in burn874/transcribe-NPU, including model.onnx, decoder_model.onnx, decoder_with_past_model.onnx and more.

  • 7 ONNX models (.onnx files)
  • 186 files, including 64 .bin, 38 .json, 23 .py, 17 .js, 8 .xml, 7 .onnx
  • Last updated 2025-07-22

ONNX files

File Size Download
models/kokoro_tts/model.onnx — Download
models/m2m100_418M/torch/decoder_model.onnx — Download
models/m2m100_418M/torch/decoder_with_past_model.onnx — Download
models/m2m100_418M/torch/encoder_model.onnx — Download
models/silero_vad/silero_vad.onnx — Download
models/silero_vad/silero_vad_16k_op15.onnx — Download
models/silero_vad/silero_vad_half.onnx — Download

How to use

These ONNX files can be loaded with ONNX Runtime. To inspect a model graph, upload the file to the Netron viewer. Download links point to the original files on Hugging Face.

README

NPU Speech Transcription Service

Local Whisper transcription / translation accelerated by Intel NPU (Core Ultra / Meteor Lake) with OpenVINO GenAI.

Prerequisites

Hardware

  • Intel NPU-capable PC (Core Ultra or newer)

System software - Windows 10/11 (PowerShell examples)

  1. Node.js LTS - front-end dev server
winget install -e --id OpenJS.NodeJS.LTS

Or grab the MSI: https://nodejs.org/en/download

  1. Conda - Python env manager
    Install Miniconda / Anaconda: https://docs.conda.io/en/latest/miniconda.html
& "C:\Windows\System32\curl.exe" -o "$env:USERPROFILE\Downloads\Miniconda3-latest-Windows-x86_64.exe" "https://repo.anaconda.com/miniconda/Miniconda3-latest-Windows-x86_64.exe"
Start-Process "$env:USERPROFILE\Downloads\Miniconda3-latest-Windows-x86_64.exe" -ArgumentList "/quiet", "/InstallAllUsers=1", "/PrependPath=1" -Wait
  1. Python
    Install Python 3.8 - 3.12. Both manual or via PowerShell works.
winget install Python.Python.3.12 --scope machine
  1. OpenVINO Runtime 2025.1 & OpenVINO GenAI 2025.1 - inference back-end
Download
& "C:\Windows\System32\curl.exe" -L -o "$env:USERPROFILE\Documents\openvino_2025.zip" "https://storage.openvinotoolkit.org/repositories/openvino/packages/2025.1/windows/openvino_toolkit_windows_2025.1.0.18503.6fec06580ab_x86_64.zip"
& "C:\Windows\System32\curl.exe" -L -o "$env:USERPROFILE\Documents\openvino_genai_2025.zip" "https://storage.openvinotoolkit.org/repositories/openvino_genai/packages/2025.1/windows/openvino_genai_windows_2025.1.0.0_x86_64.zip"
Extract
Expand-Archive $env:USERPROFILE\Documents\openvino_2025.zip -DestinationPath $env:TEMP\ov -Force
Expand-Archive $env:USERPROFILE\Documents\openvino_genai_2025.zip -DestinationPath $env:TEMP\ovg -Force
Install
$ovSubfolder = Get-ChildItem -Path "$env:TEMP\ov" -Directory | Where-Object { $_.Name -like "*openvino*" }
$ovgSubfolder = Get-ChildItem -Path "$env:TEMP\ovg" -Directory | Where-Object { $_.Name -like "*openvino_genai*" }

Move-Item "$env:TEMP\ov\$ovSubfolder" "C:\Program Files (x86)\Intel\openvino_2025" -Force
Move-Item "$env:TEMP\ovg\$ovgSubfolder" "C:\Program Files (x86)\Intel\openvino_genai_2025" -Force

Remove-Item -Path `
  "$env:USERPROFILE\Documents\openvino_2025.zip", `
  "$env:USERPROFILE\Documents\openvino_genai_2025.zip", `
  "$env:TEMP\ov", `
  "$env:TEMP\ovg" -Force -Recurse

& "C:\Program Files (x86)\Intel\openvino_2025\setupvars.ps1"
& "C:\Program Files (x86)\Intel\openvino_genai_2025\setupvars.ps1"
  1. Python wheels (inside the conda env)
conda activate
conda env create -f environment.yml
conda activate NPU

pip install --upgrade pip
pip install `
  "optimum[openvino,onnxruntime-gpu]" `
  "openvino-genai==2025.1.0" `
  "nncf" `
  "uv" `
  "fastapi" `
  "faster_whisper" `
  "huggingface" `
  "uvicorn" `
  "websockets" `
  "hf_xet" `
  "python-multipart" `
  "sentencepiece" `
  "pydub" `
  "accelerate"
  1. FFmpeg - audio decoding / resampling backend
winget install -e --id Gyan.FFmpeg
# or via Chocolatey
choco install ffmpeg --confirm

Verify:

ffmpeg -version

Initialise OpenVINO in the current shell

conda init
& "C:\Program Files (x86)\Intel\openvino_2025\setupvars.ps1"
& "C:\Program Files (x86)\Intel\openvino_genai_2025\setupvars.ps1"

Add the two lines to $PROFILE so every PowerShell session loads them automatically.

Check installation:

python -c "import openvino as ov; print(ov.Core().available_devices)"

['NPU', 'CPU'] (or similar) confirms everything is wired up.

Repository setup

git clone https://huggingface.co/burn874/transcribe-NPU NPU
cd NPU
conda env create -f environment.yml
conda activate NPU

Running

Backend only

python -m uvicorn server.main:app --host 0.0.0.0 --port 8000 --reload

Front-end only

cd frontend
npm install

# Dev server
npm start

# Production bundle
npm run build

Front-end + backend together

.\start.ps1

start.ps1 launches React (npm start) and Uvicorn, then cleans up when you exit.

API Endpoints

Path Method Description
/docs GET Swagger UI
/health GET Health check

Linux / macOS notes

  • Use the .tgz archives for Runtime & GenAI and run the accompanying .sh setup scripts.
  • Replace Move-Item with sudo mv.
  • Install Node via package manager (sudo apt install nodejs or brew install node).
  • No extra NPU driver is required on Linux.
  • FFmpeg: sudo apt install ffmpeg or brew install ffmpeg.

Troubleshooting

  • ov.Core().available_devices should list NPU. If missing, reinstall the NPU driver and reboot.
  • Run ffmpeg -encoders | findstr pcm_s16le – PCM encoders must be present.
  • Always open a new shell or re-run the setupvars scripts after installing OpenVINO.

Model info

onnx safetensors region:us
↓ 0 ♥ 0

Created: 2025-05-06

Updated: 2025-07-22

View on Hugging Face

Files (186)

.gitattributes
.gitignore
README.md
environment.yml
frontend/.gitignore
frontend/package-lock.json
frontend/package.json
frontend/public/favicon.ico
frontend/public/favicon.jpg
frontend/public/index.html
frontend/public/logo192.png
frontend/public/logo512.png
frontend/public/manifest.json
frontend/public/robots.txt
frontend/src/App.css
frontend/src/App.js
frontend/src/App.test.js
frontend/src/AppMenu.js
frontend/src/components/LanguageSettings.js
frontend/src/components/TabPanel.js
frontend/src/components/VoiceComponents.js
frontend/src/components/VoiceSelector.js
frontend/src/dashboard/Dashboard.js
frontend/src/index.css
frontend/src/index.js
frontend/src/logo.svg
frontend/src/pages/LiveTranscriptionPage.js
frontend/src/pages/ModelsPage.js
frontend/src/pages/SpeechPage.js
frontend/src/pages/TranscribeTranslatePage.js
frontend/src/reportWebVitals.js
frontend/src/utils/audioUtils.js
frontend/src/utils/commonStyles.js
frontend/src/utils/languageUtils.js
installer.ps1
models/kokoro_tts/config.json
models/kokoro_tts/config_onnx.json
models/kokoro_tts/kokoro-v1_0.pth
models/kokoro_tts/model.onnx ONNX
models/kokoro_tts/tokenizer.json
models/kokoro_tts/tokenizer_config.json
models/kokoro_tts/voices/af.bin
models/kokoro_tts/voices/af_alloy.bin
models/kokoro_tts/voices/af_aoede.bin
models/kokoro_tts/voices/af_bella.bin
models/kokoro_tts/voices/af_heart.bin
models/kokoro_tts/voices/af_jessica.bin
models/kokoro_tts/voices/af_kore.bin
models/kokoro_tts/voices/af_nicole.bin
models/kokoro_tts/voices/af_nova.bin
models/kokoro_tts/voices/af_river.bin
models/kokoro_tts/voices/af_sarah.bin
models/kokoro_tts/voices/af_sky.bin
models/kokoro_tts/voices/am_adam.bin
models/kokoro_tts/voices/am_echo.bin
models/kokoro_tts/voices/am_eric.bin
models/kokoro_tts/voices/am_fenrir.bin
models/kokoro_tts/voices/am_liam.bin
models/kokoro_tts/voices/am_michael.bin
models/kokoro_tts/voices/am_onyx.bin
models/kokoro_tts/voices/am_puck.bin
models/kokoro_tts/voices/am_santa.bin
models/kokoro_tts/voices/bf_alice.bin
models/kokoro_tts/voices/bf_emma.bin
models/kokoro_tts/voices/bf_isabella.bin
models/kokoro_tts/voices/bf_lily.bin
models/kokoro_tts/voices/bm_daniel.bin
models/kokoro_tts/voices/bm_fable.bin
models/kokoro_tts/voices/bm_george.bin
models/kokoro_tts/voices/bm_lewis.bin
models/kokoro_tts/voices/ef_dora.bin
models/kokoro_tts/voices/em_alex.bin
models/kokoro_tts/voices/em_santa.bin
models/kokoro_tts/voices/ff_siwis.bin
models/kokoro_tts/voices/hf_alpha.bin
models/kokoro_tts/voices/hf_beta.bin
models/kokoro_tts/voices/hm_omega.bin
models/kokoro_tts/voices/hm_psi.bin
models/kokoro_tts/voices/if_sara.bin
models/kokoro_tts/voices/im_nicola.bin
models/kokoro_tts/voices/jf_alpha.bin
models/kokoro_tts/voices/jf_gongitsune.bin
models/kokoro_tts/voices/jf_nezumi.bin
models/kokoro_tts/voices/jf_tebukuro.bin
models/kokoro_tts/voices/jm_kumo.bin
models/kokoro_tts/voices/pf_dora.bin
models/kokoro_tts/voices/pm_alex.bin
models/kokoro_tts/voices/pm_santa.bin
models/kokoro_tts/voices/zf_xiaobei.bin
models/kokoro_tts/voices/zf_xiaoni.bin
models/kokoro_tts/voices/zf_xiaoxiao.bin
models/kokoro_tts/voices/zf_xiaoyi.bin
models/kokoro_tts/voices/zm_yunjian.bin
models/kokoro_tts/voices/zm_yunxi.bin
models/kokoro_tts/voices/zm_yunxia.bin
models/kokoro_tts/voices/zm_yunyang.bin
models/m2m100_418M/torch/README.md
models/m2m100_418M/torch/added_tokens.json
models/m2m100_418M/torch/config.json
models/m2m100_418M/torch/decoder_model.onnx ONNX
models/m2m100_418M/torch/decoder_with_past_model.onnx ONNX
models/m2m100_418M/torch/encoder_model.onnx ONNX
models/m2m100_418M/torch/generation_config.json
models/m2m100_418M/torch/sentencepiece.bpe.model
models/m2m100_418M/torch/special_tokens_map.json
models/m2m100_418M/torch/tokenizer_config.json
models/m2m100_418M/torch/vocab.json
models/m2m100_418M/vino/added_tokens.json
models/m2m100_418M/vino/config.json
models/m2m100_418M/vino/generation_config.json
models/m2m100_418M/vino/openvino_decoder_model.bin
models/m2m100_418M/vino/openvino_decoder_model.xml
models/m2m100_418M/vino/openvino_decoder_with_past_model.bin
models/m2m100_418M/vino/openvino_decoder_with_past_model.xml
models/m2m100_418M/vino/openvino_encoder_model.bin
models/m2m100_418M/vino/openvino_encoder_model.xml
models/m2m100_418M/vino/sentencepiece.bpe.model
models/m2m100_418M/vino/special_tokens_map.json
models/m2m100_418M/vino/tokenizer_config.json
models/m2m100_418M/vino/vocab.json
models/silero_vad/silero_vad.jit
models/silero_vad/silero_vad.onnx ONNX
models/silero_vad/silero_vad_16k_op15.onnx ONNX
models/silero_vad/silero_vad_half.onnx ONNX
models/whisper-small/torch/added_tokens.json
models/whisper-small/torch/config.json
models/whisper-small/torch/flax_model.msgpack
models/whisper-small/torch/generation_config.json
models/whisper-small/torch/merges.txt
models/whisper-small/torch/model.safetensors
models/whisper-small/torch/normalizer.json
models/whisper-small/torch/preprocessor_config.json
models/whisper-small/torch/pytorch_model.bin
models/whisper-small/torch/special_tokens_map.json
models/whisper-small/torch/tf_model.h5
models/whisper-small/torch/tokenizer.json
models/whisper-small/torch/tokenizer_config.json
models/whisper-small/torch/vocab.json
models/whisper-small/vino/added_tokens.json
models/whisper-small/vino/config.json
models/whisper-small/vino/generation_config.json
models/whisper-small/vino/merges.txt
models/whisper-small/vino/normalizer.json
models/whisper-small/vino/openvino_decoder_model.bin
models/whisper-small/vino/openvino_decoder_model.xml
models/whisper-small/vino/openvino_decoder_with_past_model.bin
models/whisper-small/vino/openvino_decoder_with_past_model.xml
models/whisper-small/vino/openvino_detokenizer.bin
models/whisper-small/vino/openvino_detokenizer.xml
models/whisper-small/vino/openvino_encoder_model.bin
models/whisper-small/vino/openvino_encoder_model.xml
models/whisper-small/vino/openvino_tokenizer.bin
models/whisper-small/vino/openvino_tokenizer.xml
models/whisper-small/vino/preprocessor_config.json
models/whisper-small/vino/special_tokens_map.json
models/whisper-small/vino/tokenizer.json
models/whisper-small/vino/tokenizer_config.json
models/whisper-small/vino/vocab.json
server/config.json
server/config.py
server/device.py
server/main.py
server/pipelines/KokoroTTS.py
server/pipelines/OVWhisperTranscriber.py
server/pipelines/VAD.py
server/pipelines/Whisper.py
server/routers/list_models.py
server/routers/transcribe.py
server/routers/transcribe_ws.py
server/routers/translate.py
server/routers/tts.py
server/services/transcribe_service.py
server/services/translate_service.py
start.ps1
start.sh
test.mp3
test.py
test.wav
tests/run_all_tests.py
tests/test_models_api.py
tests/test_server_functions.py
tests/test_transcribe_service.py
tests/test_transcribe_ws.py
tests/test_transcription_api.py
tests/test_translation_api.py
tests/test_tts_api.py

Related models

More from burn874

More models

No task type declared.

Popular models