Anima Base Local Inpainting & Masked Redraw Workflow

This workflow is designed for Anima Base local inpainting and masked redraw. Its main purpose is to let creators upload an existing anime-style image, paint or provide a mask over the area that needs to be changed, and then use Anima Base to regenerate only the selected region while keeping the rest of the image as stable as possible.

Unlike a full image-to-image workflow, this setup is focused on partial correction. It is useful when the whole image is already good, but one part needs repair or replacement: hands, fingers, face details, clothing, hair, accessories, background defects, small anatomy errors, unwanted objects, or local composition problems. Instead of regenerating the entire picture and risking identity drift, this workflow uses mask-based latent editing so the correction stays concentrated inside the selected area.

The workflow uses anima_baseV10.safetensors as the core UNet model, qwen_3_06b_base.safetensors as the text encoder route, and qwen_image_vae.safetensors as the VAE. It includes LoadImage, image_scale_pixel_v2, GrowMaskWithBlur, VAEEncode, SetLatentNoiseMask, DifferentialDiffusion, CLIPTextEncode, ClownsharKSampler_Beta, VAEDecode, PreviewImage, and SaveImage. This creates a compact but practical local-redraw pipeline for anime image repair.

The key part of this workflow is the mask route. The uploaded image provides both the image and mask. The image is scaled through image_scale_pixel_v2, while the mask is processed through GrowMaskWithBlur. This mask expansion and blur step is important because hard mask edges often create visible seams. By expanding and softening the mask, the regenerated area can blend more naturally into the surrounding pixels.

SetLatentNoiseMask is another important node. It tells the sampler which part of the latent image should receive noise and be regenerated. This is the basic logic behind local inpainting: the masked area is allowed to change, while the unmasked area should remain close to the original. DifferentialDiffusion helps make the masked redraw process more controlled, which is useful when the creator wants local repair rather than full-scene transformation.

The positive prompt in the workflow focuses on anime-style high-quality output, safe image generation, high resolution, luxury fashion illustration, and a dancing young adult female character. The negative prompt suppresses low quality, blur, bad anatomy, bad hands, extra fingers, malformed limbs, duplicate elements, cropped results, deformed faces, poorly drawn eyes, text, and watermark artifacts. This makes the workflow suitable for common anime-image cleanup tasks, especially hand repair and localized visual correction.

The sampling stage uses ClownsharKSampler_Beta with a relatively direct inpainting configuration. The workflow is designed to regenerate the masked area with enough strength to fix local problems while still preserving the original image outside the mask. The final result is decoded through VAEDecode and displayed through PreviewImage, with SaveImage available for exporting the finished repair.

This workflow is ideal for Anima Base local redraw, anime character repair, hand and face correction, masked outfit editing, small object removal, background cleanup, accessory adjustment, and localized image enhancement. It is especially useful for creators who want to polish AI-generated anime illustrations without destroying the entire image. If you want to see how mask input, blurred mask expansion, latent inpainting, DifferentialDiffusion, Anima Base sampling, and final local repair output work together, watch the full tutorial from the YouTube link above.

⚙️ Try the Workflow Online

👉 Workflow: https://www.runninghub.cn/post/2055109942587215873?inviteCode=rh-v1111

Open the link above to run the workflow directly online and view the generation results in real time.

If the results meet your expectations, you can also deploy it locally for further customization.

🎁 Fan Benefits: Register now to get 1000 points, plus 100 daily login points — enjoy 4090-level performance and 48 GB of powerful compute!

📺 Bilibili Updates (Mainland China & Asia-Pacific)

If you are in Mainland China or the Asia-Pacific region, you can watch the video below for workflow demos and a detailed creative breakdown.

📺 Bilibili Video: https://www.bilibili.com/video/BV1ci5q65EGQ/

I will continue updating model resources on Quark Drive:

👉 https://pan.quark.cn/s/20c6f6f8d87b

These resources are mainly prepared for local users, making creation and learning more convenient.

⚙️ 在线体验工作流

👉 工作流： https://www.runninghub.cn/post/2055109942587215873?inviteCode=rh-v1111

打开上方链接即可直接运行该工作流，实时查看生成效果。

如果觉得效果理想，你也可以在本地进行自定义部署。

🎁 粉丝福利： 注册即送 1000 积分，每日登录 100 积分，畅玩 4090 体验 48 G 超级性能！

📺 Bilibili 更新（中国大陆及南亚太地区）

如果你在中国大陆或南亚太地区，可以通过下方视频查看该工作流的实测效果与构思讲解。

📺 B站视频： https://www.bilibili.com/video/BV1ci5q65EGQ/

我会在 夸克网盘 持续更新模型资源：

👉 https://pan.quark.cn/s/20c6f6f8d87b

这些资源主要面向本地用户，方便进行创作与学习。