RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2078868414222913537?inviteCode=rh-v1111
Krea 2 Advanced Cleanup - RunningHub Online Batch (Structure LoRA + Reverse Prompt)
Purpose: the RunningHub-hosted batch variant. Upload a ZIP of images, tell the graph how many images to process, and it loops: each image is reverse-prompted, depth-encoded, rebuilt through the structure stages and upscaled, then saved one by one on the platform.
Graph (as saved)
66 nodes / 109 links / 10 groups.
- Model: UNETLoader
krea2_turbo_fp8.safetensors; CLIPLoaderqwen3vl_4b_fp8_scaled.safetensors(typekrea2); VAELoaderqwen_image_HDR_vae_fp32_comfy.safetensors. - Structure LoRA: one
Krea2ControlLoRALoaderondepth-control-lora.safetensorsat strength 1.0. - Platform I/O:
RHUploadZip(ZIP upload),RHBatchImages:with an image range such as 0-9,RH_ExtractImage, and a for-loop start/end pair — this is what makes it an online batch rather than a local folder loop. - Three depth stages: three
Krea2ControlImageEncode→Krea2ControlApplypairs. - Sampling: three KSamplers at 8 steps / CFG 1 / euler / simple with denoise 0.95, 0.35 and 0.25; two LatentUpscale nodes (bislerp, 1024) saved disabled; final 4x-UltraSharp upscale.
- Reverse prompt: Qwen3VLProcessor (open) and RH_LLMAPI_Pro_Node (closed, GLM-4.6V label);
ConditioningKrea2Rebalanceactive. - Output: SaveImage prefix
Krea2_线上批量洗图.
How to use
Upload a ZIP that contains only jpg/jpeg/png/webp images and enter the exact image count — the graph generates the 0..(n-1) range from it. Per the saved note, each image gets its own reverse prompt, own width/height and own depth extraction; results are not carried across images. Outside RunningHub the RH nodes will not resolve.
Limits and not-verified items
No inference was run for this listing: no speed, VRAM, batch throughput, cost or quality claim is made. RunningHub node behaviour is platform-side. Weights, LoRA, depth model, upscaler and VLM routes are not bundled; check their licenses separately. 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, LoRA, depth model and cover rights separately. No blanket commercial or redistribution license is asserted.
Description
RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2078868414222913537?inviteCode=rh-v1111
Krea 2 Advanced Cleanup - RunningHub Online Batch (Structure LoRA + Reverse Prompt)Purpose: the RunningHub-hosted batch variant. Upload a ZIP of images, tell the graph how many images to process, and it loops: each image is reverse-prompted, depth-encoded, rebuilt through the structure stages and upscaled, then saved one by one on the platform.
Graph (as saved)66 nodes / 109 links / 10 groups.
- Model: UNETLoader
krea2_turbo_fp8.safetensors; CLIPLoaderqwen3vl_4b_fp8_scaled.safetensors(typekrea2); VAELoaderqwen_image_HDR_vae_fp32_comfy.safetensors. - Structure LoRA: one
Krea2ControlLoRALoaderondepth-control-lora.safetensorsat strength 1.0. - Platform I/O:
RHUploadZip(ZIP upload),RHBatchImages:with an image range such as 0-9,RH_ExtractImage, and a for-loop start/end pair — this is what makes it an online batch rather than a local folder loop. - Three depth stages: three
Krea2ControlImageEncode→Krea2ControlApplypairs. - Sampling: three KSamplers at 8 steps / CFG 1 / euler / simple with denoise 0.95, 0.35 and 0.25; two LatentUpscale nodes (bislerp, 1024) saved disabled; final 4x-UltraSharp upscale.
- Reverse prompt: Qwen3VLProcessor (open) and RH_LLMAPI_Pro_Node (closed, GLM-4.6V label);
ConditioningKrea2Rebalanceactive. - Output: SaveImage prefix
Krea2_线上批量洗图.
Upload a ZIP that contains only jpg/jpeg/png/webp images and enter the exact image count — the graph generates the 0..(n-1) range from it. Per the saved note, each image gets its own reverse prompt, own width/height and own depth extraction; results are not carried across images. Outside RunningHub the RH nodes will not resolve.
Limits and not-verified itemsNo inference was run for this listing: no speed, VRAM, batch throughput, cost or quality claim is made. RunningHub node behaviour is platform-side. Weights, LoRA, depth model, upscaler and VLM routes are not bundled; check their licenses separately. 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, LoRA, depth model and cover rights separately. No blanket commercial or redistribution license is asserted.