Fine-edge priority
Built for cases where thin structures, hair, spokes, or complex silhouettes need more attention than the fast path.
When the fast mask is not enough.
Use the heavier edge-quality model.
High Quality mode uses BiRefNet Lite on WebGPU for difficult silhouettes and automatically falls back to Universal AI when the device cannot run it.
Private AI photo editor
Drop a photo. Keep the subject. Transform the background.
Your image is processed on this device — it is not uploaded to our server.
Automatic fallback
If WebGPU or the heavy model cannot initialize, the editor switches to Universal AI instead of stranding the edit.
Manual final control
Preserve the model's soft alpha matte, adjust edge contrast and softness, then Restore or Erase details manually.
How it works
Useful output, without a cloud editing queue.
High Quality AI is intentionally not the default. Its fp16 browser graph is much larger than the fast models, so making every visitor download it would damage first-use performance.
Use it when edge accuracy is worth the extra model load. For normal images, Universal AI is the better default; for straightforward portraits, Portrait AI remains the fastest specialist.