← Models

SOUND / Whisper

Whisper

Jetson Orin Nano

Robust speech recognition model via OpenAI Whisper.

245 downloads
Size
~49MB+162.6MB
Memory
2GB+
Precision
INT8

Choose the device you're using, the set up guide and documentation will update accordingly.

Getting Started

Deploy
# Docker command for rknn on jetson-orin-nano
sudo docker run -it --rm \
  <image>:latest \
  <serve> whisper

REST API

Use the REST API to run inference. Copy the commands below.

Curl
curl -X POST "http://127.0.0.1:8000/api/models/whisper/task" -F "file=@/home/user/audio/long_podcast.wav" -F "language=zh"
Python
import requests
import json
resp = requests.post(
    "http://127.0.0.1:8000/api/models/whisper/predict",
    files={"file": open("/home/user/audio/test_en.wav", "rb")},
    data={"language": "en"},
    timeout=30
)
result = resp.json()
print(json.dumps(result, indent=2, ensure_ascii=False))

Model Details

reComputer Jetson

Jetson One-Command Deployment

Run the command in the Getting Started block directly on a Jetson or reComputer Jetson device.

bash
python3 -m pip install -U jetson-examples && reComputer run <demo-name>

The command installs or upgrades the jetson-examples Python package from PyPI, then runs the selected demo script through the reComputer CLI.

Requirements

  • NVIDIA Jetson or reComputer Jetson device.
  • JetPack/L4T version supported by the selected demo.
  • Internet access for PyPI packages, Docker images, model weights, or external assets used by the demo.
  • Docker and NVIDIA container runtime when the selected demo uses containers.
  • Enough disk and memory for the selected demo. The model card lists the main resource requirement when it is known from the demo configuration.

Workflow

  1. Copy the deployment command from the Run dialog or detail page.
  2. Run it in a terminal on the target Jetson device.
  3. Follow any prompts printed by the demo script.
  4. Open the URL shown in the terminal if the demo starts a web UI or API service.

Notes

  • reComputer run <demo-name> runs the demo's init.sh first when available, then starts run.sh.
  • To inspect all packaged demos on the device, run:
bash
reComputer list
  • To remove demo data when a cleanup script exists, run:
bash
reComputer clean <demo-name>