This workflow is designed for WAI-ANIMA tiled image upscaling and detail refinement. Its main purpose is to enlarge anime-style images while preserving character structure, line quality, local details, and overall composition. Instead of simply running a single global upscale pass, this workflow first enlarges the source image, then divides the result into smaller tiles, refines each tile through an Anima-based generation route, and finally reconstructs the full image with better local detail.
The workflow uses waiANIMA_v10 as the main UNet model, qwen_3_06b_base as the text encoder, and qwen_image_vae as the VAE. It also includes a LoRA route, with Gundam_RX78_Flux loaded in the example setup, showing that the workflow can be adapted for specific character, mecha, or style enhancement tasks. For the initial enlargement stage, it uses RealESRGAN_x4plus_anime_6B, which is especially suitable for anime-style images with clean linework and illustrated textures.
The key structure is the tiled processing pipeline. The input image is first upscaled, then scaled to a larger total pixel target. After that, TTP_Tile_image_size calculates the proper tile layout, and TTP_Image_Tile_Batch splits the image into tile batches. Each tile is encoded through the VAE, processed with the Anima model, and later decoded through tiled VAE decoding. Finally, TTP_Image_Assy reconstructs the processed tiles back into a complete image.
This tiled method is useful because large anime images often fail when processed in one pass. A single-pass upscale may produce soft details, broken linework, noisy textures, face degradation, or inconsistent local areas. By processing the image in tiles, the workflow gives each region more dedicated generation attention. This helps improve hair strands, clothing folds, armor details, background elements, facial features, and fine illustration textures.
The workflow also includes WD14 tag extraction. The tiled image regions can be analyzed with a tagger, then the extracted visual tags are combined with an artist/style prompt through JWStringConcat. In the example setup, the prompt includes “@sakimichan + @artgerm,” showing how the workflow can guide local refinement toward a specific polished illustration style. This is useful when creators want the upscale to do more than enlarge pixels: they want the image to gain a cleaner, more stylized, more finished look.
The sampler section uses ClownsharKSampler_Beta with a relatively low denoise value, which is important for controlled enhancement. A high denoise value could redraw the image too aggressively, while a lower denoise setting helps preserve the original composition and character identity. This makes the workflow more suitable for refinement, enhancement, and upscale restoration rather than full image redesign.
The workflow also includes comparison output through an image comparer node. This makes it easy to inspect the before-and-after difference, check whether the tiled reconstruction is seamless, and decide whether the upscale pass improved the final result without damaging the original design.
This workflow is ideal for anime illustration upscaling, character art enhancement, mecha image refinement, Civitai preview polishing, social media cover preparation, and high-resolution local detail repair. If you want to see how WAI-ANIMA, RealESRGAN anime upscale, WD14 tagging, tiled processing, local refinement, and final image reconstruction work together, watch the full tutorial from the YouTube link above.
⚙️ Try the Workflow Online
👉 Workflow: https://www.runninghub.ai/post/2046194195202838530/?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/BV1q7drB7Ecp/
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.ai/post/2046194195202838530/?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/BV1q7drB7Ecp/
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。