Google Research proposes Diffusion Controller for image generation control
What happened
On 30 September 2026, Google Research released Diffusion Controller, a framework that recasts diffusion-model denoising as a continuous control problem. A lightweight steering-damper network dynamically adjusts generation trajectories while keeping the base model frozen. The report says its fully unlocked version achieved a 90% win rate over the baseline on Stable Diffusion v1.4 using HPS-v2 evaluation, and supports customised control of closed models whose internal weights are inaccessible. The release and evaluation results currently come from Google Research’s primary report, with no independent verification or further developments from other sources.
Written by AI from the reports · updated 22 h ago
Timeline
Follow the reports to see the event from each side.
- Google ResearchHow Diffusion Controller unifies and simplifies AI image generation
Google Research proposed Diffusion Controller, reframing diffusion-model denoising as a continuous control problem. A lightweight steering-damper network dynamically adjusts generation trajectories while the base model remains frozen. Evaluated on Stable Diffusion v1.4 using HPS-v2, its fully unlocked version achieved a 90% win rate against the baseline, and it supports customised control of closed models without access to internal weights.
Heat of this event
Current heat 3·Peak in the comparable range 5（Oct 1 01:00）·Change over 24 hours in the comparable range –
The trend compares only the same accounts observed without a break, so its range can be narrower than the current heat count. Move the pointer or tap the chart to see each hour; with a keyboard, use the left and right arrow keys.