FLUX 3 Image Klein distilled editing workflow. Same bounding-box and multi-reference editing as the base graph, but on the distilled checkpoint so iteration stays fast.
Main features:
- Built on the distilled FLUX 3 Image Klein checkpoint (flux3_image_klein_sd_int8_convrot + qwen3vl_4b text encoder + flux2 VAE)
- 16 steps, no classifier-free guidance, euler/simple, denoise 1.0 - no guidance needed on the distilled path
- Bounding-box editing via CreateBoundingBoxes + Flux3LayoutPrompt, addressed as <object_1>, <object_2> in the caption
- BatchImagesNode + ReferenceMultiLatent accept up to 10 reference images, addressed as ref_image_0, ref_image_1, ...
- ComfySwitchNode toggles edit variants without duplicating the graph; PrimitiveStringMultiline keeps long captions tidy
- Prompt upsampler included but bypassable when you bring your own detailed caption
Suggested workflow:
1. Load the image and draw boxes around the regions to edit.
2. Write the caption as a full description of the wanted final frame, naming each box.
3. Run the graph. Because this is the distilled path, run several seeds first and only then compare against the base editing graph for the final pick.
Dependencies:
- flux3_image_klein_sd_int8_convrot.safetensors (distilled UNET)
- qwen3.5_9b_flux3_prompt_upsampler_w4a8.safetensors (prompt upsampler)
- qwen3vl_4b_int8_convrot.safetensors (text encoder / CLIP)
- flux2-vae.safetensors (VAE)
- BatchImagesNode, ComfySwitchNode, CreateBoundingBoxes, Flux3LayoutPrompt, PrimitiveStringMultiline, ReferenceMultiLatent, TextGenerate (custom nodes)
RunningHub workflow page: https://www.runninghub.ai/zh-cn/user-center/1846601088051884033/webapp?inviteCode=rh-v1111
Bilibili tutorial video: https://www.bilibili.com/video/BV1btJb6LEGS/
RunningHub registration with invite code rh-v1111: https://www.runninghub.ai/?inviteCode=rh-v1111