UniSlider adds a perceptual slider to control how much an AI image edit changes
CSProfKGD · x · 2026-10-17
UniSlider, a project from an Adobe Research internship by serra9lozano, tackles a long-standing pain point in image editing: text prompts say what to edit but not how much.
The tool introduces a slider whose value is perceptually proportional to the magnitude of the change in the image, letting users continuously and quantitatively control edit intensity instead of iterating on prompts blindly.
- Project page and paper are publicly available
- Result of the author's internship at Adobe Research
More from Multimodal
- ElevenLabs ships Instant Soundtracks: auto-scored video with timestamped SFX — lmoroney · 2026-10-17
- OpenMOSS open-sources MOSS-Image, a 5B DiT model focused on text rendering — KonoDIOda_666 · 2026-10-17
- Fully local AI video generation with Claude as the director — No_Mud2447 · 2026-10-17
- Generate AI video in 480p then upscale? Debate over credit-saving trick — adrianscottcom · 2026-10-17
- Voice-prompting industrial techno: describing clacking keyboard sounds to an AI music model — michellechen · 2026-10-17
- MiniMax-H3 plus ORB360 LoRA demoed making comics with freely movable camera angles — MiniMax_AI · 2026-10-17