Skip to main content

​
GPU Acceleration

Echophrase uses ONNX Runtime’s automatic device selection to pick the fastest available hardware for transcription at runtime. You don’t need to configure anything — the app detects your GPU on first launch and uses it automatically.

​
Supported Hardware

​
CUDA (optional, Windows + NVIDIA)

DirectML is the default on Windows because it ships with the OS and works on every modern GPU regardless of vendor - no driver/toolkit version juggling. If you already have the CUDA Toolkit 13.0+ and cuDNN 9.19+ installed, you can opt in to the CUDA execution provider from Settings -> GPU Acceleration. Echophrase downloads the CUDA runtime as a separate plugin (~650 MB) and activates it after a restart. It is entirely optional: leaving it alone keeps you on DirectML. The published Linux build is CPU-only.

​
Verifying GPU Usage

  1. Open Echophrase Settings
  2. Look at GPU Acceleration — the top line shows the active device (for example: “Active: Using DirectML (NVIDIA GeForce RTX 3080)”)
  3. Transcribe a clip and confirm the transcription is fast
If the active device says “Using CPU” but you have a DirectX 12 GPU installed, see GPU Troubleshooting.

​
Troubleshooting

If GPU acceleration isn’t working:
  • Update your GPU drivers to the latest from the vendor (NVIDIA / AMD / Intel)
  • Confirm your GPU supports DirectX 12 (most cards from 2015 onwards do)
  • Check GPU Troubleshooting for platform-specific help