Sogni: Learn logo

Advanced Features

Pocket's default flow is text-to-image: type a prompt, pick a style, render. Advanced Features are the tools you reach for when you want more control over the result — masking specific regions, anchoring to a reference image, or guiding generation with sketches, poses, depth, or faces.

Each one composes with the others: you can stack a Guide Image, a ControlNet pose, and a Paintover mask on the same render.

#Paintover

Paintover lets you paint a mask directly on an image and regenerate only the masked regions. Use it to change hair color while keeping the face, swap clothing, remove a blemish, add a hat, or repaint a background — anything where the rest of the image should stay put. It's Pocket's mobile take on inpainting, with a brush-driven UI sized for touch.

#Guide Image

Guide Image is image-to-image: feed in a reference photo, and the model uses it as a starting point for composition, color, and mood. A Strength slider controls how closely the result hugs the reference versus how much the model reinterprets it. Good for style transfer, variations on an existing shot, and locking in a mood without prescribing exact structure.

#ControlNet

ControlNet adds structural guidance from feature maps — sketches, body poses, depth maps, and face landmarks — rather than the raw pixels of the reference. Use it when you want to dictate "this exact pose" or "this exact composition" while still letting the prompt and style drive everything else. Pocket exposes four ControlNet flavors: Sketch a Masterpiece, Face and Pose Capture, Face Transfer, and Depth Map.

#When to reach for which

  • Same subject, different look → Guide Image at lower strength.
  • Same image, change one region → Paintover.
  • Specific pose, sketch, or face → ControlNet.
  • All three together → Yes, you can stack them. Use Guide Image for the look, ControlNet for the structure, and Paintover to polish a region after.

#See also