Automating Stable Diffusion’s Boring Batch Work

What it does

This is glue code that adds two scripts to AUTOMATIC1111’s Stable Diffusion WebUI. Enhanced img2img batch-processes folders of images for img2img and inpainting, while Multi-frame rendering chains frames together to keep sequences consistent. Both can pull prompts from text files or CSVs so you don’t have to retype them for every image.

The interesting bit

Multi-frame rendering is explicitly designed to pair with ControlNet and offers reference-frame modes like FirstGen, OriginalImg, and Historical to reduce color fading between outputs. Enhanced img2img can also hijack a PNG’s alpha channel as an inpainting mask, which skips the need for a separate mask directory.

Key highlights

• Batch-process directories with optional mask folders or use a PNG alpha channel as the mask
• Multi-frame rendering with reference-frame options (FirstGen, OriginalImg, Historical) and configurable loopback sources
• Prompt automation via DeepDanbooru/CLIP interrogation, text files, or CSV lists
• Optional color correction and automatic seed incrementing between frames
• Zoom-and-crop on masked areas to improve results when the masked region is small

Caveats

• The author admits the contextual-information tag feature improves accuracy only “maybe”
• Multi-frame rendering expects ControlNet to be useful
• Documentation beyond the README is video-only and in Chinese

Verdict

Grab it if you’re batch-processing images or trying to keep Stable Diffusion outputs consistent across frames. Pure one-off prompt explorers can skip it.