Features
Portrait blur (AI bokeh)
Blur the background while keeping your subject sharp using AI depth estimation.
How it works
Portrait blur uses an AI depth estimation model (MiDaS) to estimate how far each pixel is from the camera. Pixels estimated as far away (background) receive more blur; pixels close to the camera (subject) receive no blur.
This is different from background removal — the subject and background are both retained, but the background is blurred to create a shallow depth-of-field effect.
Controls
- Blur strength: How strong the blur effect is at maximum depth. Higher values create stronger bokeh.
- Depth threshold: How aggressively "near" and "far" regions are separated. Increase if background areas are bleeding into the subject.
- Blur radius: The pixel radius of the Gaussian blur kernel. Larger values = smoother, more blended bokeh.
Tips for best results
- Works best on images with a clear subject in front of a distinct background
- Portraits, product shots, and pet photos work particularly well
- Very flat scenes (a face looking at a wall from 10cm) give poor depth separation
- If the subject edges are blurring, lower the depth threshold
Limitations
- Depth estimation is approximate — it can misjudge depth on complex scenes
- Transparent or reflective surfaces confuse the model
- Very fine details (loose hair, thin glasses frames) may get some blur at the edges