AI Tools
OpenGPEX integrates powerful AI capabilities that run directly in your browser — no cloud processing required for core features.
Overview
The AI Tools are accessible from the AI Tools drawer in the right sidebar. Three core AI capabilities are available:
| Tool | Purpose | Processing |
|---|---|---|
| Background Removal | Isolate subjects from backgrounds | Client-side inference |
| Upscale (Super Resolution) | Enlarge images with AI detail enhancement | Client-side inference |
| Segmentation (SAM) | Identify and separate objects/regions | Client-side inference |
All three tools run on your device — inference uses WebGPU where available with a WASM fallback (ONNX Runtime / transformers.js) — your images never leave your browser.
AI Background Removal
What It Does
Automatically detects the main subject in your image and removes the background, creating a transparent mask.
How to Use
- Select the layer containing your image
- Open AI Tools drawer → Background Removal
- Click the removal button
- Wait for AI processing (a few seconds depending on image size)
- The foreground subject is detected and applied as a clip selection — refine or delete the background from there
Models Available
- RMBG 1.4: Fast, general-purpose background removal
- InSPyReNet Ultra: High-quality edge detection for complex subjects (hair, fur, etc.)
Tips
- Works best with clearly defined subjects (people, objects, animals)
- For complex edges, use the Eraser tool to refine after AI removal
- Press Tab to switch to Restore tool to bring back any incorrectly removed areas
AI Upscale (Super Resolution)
What It Does
Enlarges your image by 2× or 4× while using AI to generate realistic detail that doesn't exist in the original.
How to Use
- Select the layer you want to upscale
- Open AI Tools drawer → Upscale
- Choose a model — the scale factor comes from the model: 2× AnimeSharp V4 Fast (2×) or 4× Real-ESRGAN General (4×)
- Choose the output mode: a new creation (default) or replace the current frame
- Click to start processing
When to Use
- Low-resolution source images that need to be printed larger
- Cropped images where you've lost resolution
- Thumbnail-size images that need to be full-sized
Limitations
- Processing time increases significantly with image size
- 4× upscale of a large image may take considerable time
- Best results on photos; less predictable on text or vector graphics
AI Segmentation (SAM 2.1)
What It Does
Interactive object segmentation using SAM 2.1 (Segment Anything Model). Click or draw a box to select any object with pixel-perfect precision.
How to Use
- Select the SAM tool in Clip mode
- Point prompt: Click on the object you want to select
- Box prompt: Drag a rectangle around the target object
- Auto segment: Run full-image segmentation to detect all objects
- Use the resulting selection for Peel, Stamp, masking, or deletion
Use Cases
- Isolate specific objects (car, person, sky, building, etc.)
- Replace traditional magnetic lasso / quick select workflows
- Selective editing — adjust only specific objects in the scene
Privacy & Security
- Core AI tools (Background Removal, Upscale, Segmentation) run entirely in your browser
- Images are processed locally (WebGPU when available, WASM fallback via ONNX Runtime)
- No data is sent to external servers for these operations
- ComfyUI Bridge connects only to servers you configure
Next Steps
- Masks & Non-Destructive Editing — Refine AI-generated masks
- Color Adjustments — Enhance AI-processed images
- Export & File Formats — Save your AI-enhanced work
Last updated: 2026-07-30