RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2089668746540572673?inviteCode=rh-v1111

Anima 2.9B Three-Stage High-Quality Render

Purpose: a three-pass Anima 2.9B render. The base image is generated, then two latent upscales with low-denoise passes progressively resolve detail, so the final image is built through three sampling stages rather than one.

Graph (as saved)

16 nodes / 23 links.

• UNETLoader Anima-2.9B-preview-v1.safetensors; CLIPLoader qwen_3_06b_base.safetensors (type stable_diffusion); VAELoaderKJ qwen_image_vae.safetensors.
• Base canvas: EmptyLatentImage 1024x1024 + FluxResolutionNode at 1.0 MP.
• Three KSamplers, all 28 steps / CFG 5 / res_multistep / linear_quadratic: stage 1 denoise 1, stage 2 denoise 0.35, stage 3 denoise 0.3.
• Between stages: two LatentUpscaleBy nodes, nearest-exact at 1.25x each.
• SaveImage prefix Anima_2.9B/three_stage_hq; a PreviewImage is kept in the graph for the intermediate stage.

How to use

The two 0.35 / 0.3 denoise values are the main tuning surface: raise them if the upscaled stages look soft, lower them if the model starts repainting the subject. Changing the two 1.25x factors changes the final size, so re-check the output resolution after editing them.

Limits and not-verified items

No inference was run for this listing: no speed, VRAM or quality claim is made; values are quoted from the saved nodes. Weights, VAE, encoder and demo images are not bundled. The cover image is a promotional still, not a frame produced by this graph.

Tutorial: YouTube · Bilibili. Workflows supplied through AI-KSK; verify workflow, node, model, encoder, image and cover rights separately. No blanket commercial or redistribution license is asserted.