<div data-youtube-video="https://youtu.be/aCEuir7zy-w"><iframe src="https://www.youtube.com/embed/aCEuir7zy-w" width="853" height="480" allowfullscreen></iframe></div> 
RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2080681456040501249?inviteCode=rh-v1111
<h2>Krea 2 Cleanup 2.6 - Single-Image (Three-Stage Depth Control + Presets)</h2>
Purpose: the single-image variant of the Krea 2 cleanup set. One image in, the same three depth-conditioned stages (high-noise shaping, mid-noise structure hold, low-sigma convergence), RAW + Turbo base models and a 4x upscale pass, with one preset number controlling how far the edit is allowed to move.
<h2>Graph (as saved)</h2>
89 nodes / 149 links / 38 groups.
<ul>
<li>Depth chain: three <code>Krea2ControlImageEncode</code> stages each feeding <code>Krea2ControlApply</code> and a <code>Krea2ControlLoRALoader</code> on <code>depth-control-lora.safetensors</code> at 1.00 / 0.85 / 0.65 (preset-driven).</li>
<li>Three more <code>LoraLoaderModelOnly</code> slots on the same file are saved at strength 0 - disabled slots, not active LoRAs.</li>
<li>Models: <code>krea2_raw.safetensors</code> high-noise stage, <code>krea2_turbo_fp8.safetensors</code> mid/low. CLIP <code>qwen3vl_4b_fp8_scaled.safetensors</code> (type <code>krea2</code>), VAE <code>qwen_image_HDR_vae_fp32_comfy.safetensors</code>.</li>
<li>Single input: one <code>LoadImage</code>; depth via DA3 nodes; <code>ConditioningKrea2Rebalance</code> saved bypassed in this variant.</li>
<li>Samplers: three <code>SamplerCustomAdvanced</code> with one shared <code>SamplerEulerAncestral</code> (1, 1); shared RAW/Turbo schedule 36 steps denoise 0.92; Turbo upscale refine 10 steps denoise 0.35; active intermediate KSampler 10 steps, CFG 1, euler_ancestral, simple, denoise 0.40. Legacy KSamplers saved bypassed (8j0.90, 10j0.30).</li>
<li>Upscale: 4x-UltraSharp + <code>ImageUpscaleWithModel</code>; two <code>LatentUpscale</code> nodes. SaveImage prefix <code>Krea2_单图洗图_2.4_RAWTurbo</code>.</li>
<li>Preset dial: 0 = tightly controlled / faithful, 1 = balanced (default), 2 = diverging high-noise repaint; drives three denoise values, Split checkpoints, RAW CFG and the depth-LoRA strength trio together (preset 2 sets depth LoRA to 0).</li>
</ul>
<h2>How to use</h2>
Load one image, pick the preset, and write or extract the prompt. The graph's CLIP node accepts the final prompt from the JSON field; the LLM/JSON caption branch is shared with the batch variants. Leave the depth-LoRA trio at preset values unless you deliberately override it.
<h2>Limits and not-verified items</h2>
No inference was run for this listing: no speed, VRAM, fidelity or upscale-quality claim is made. The saved nodes and notes are the author's configuration, not measured results. Weights, LoRA, depth model and input media are not bundled; check their licenses separately. The cover image is a promotional still, not a frame produced by this graph.
Tutorial: <a href="https://www.youtube.com/watch?v=aCEuir7zy-w" rel="ugc">YouTube</a> · <a href="https://www.bilibili.com/video/BV1GBgi6JEwL" rel="ugc">Bilibili</a>. Workflows supplied through AI-KSK; verify workflow, node, model, LoRA, input-image and cover rights separately. No blanket commercial or redistribution license is asserted.

Krea 2 Cleanup 2.6 - Single-Image (Three-Stage Depth Control + Presets)

<div data-youtube-video="https://youtu.be/aCEuir7zy-w"></div> 
RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2080681456040501249?inviteCode=rh-v1111
Krea 2 Cleanup 2.6 - Single-Image (Three-Stage Depth Control + Presets)
Purpose: the single-image variant of the Krea 2 cleanup set. One image in, the same three depth-conditioned stages (high-noise shaping, mid-noise structure hold, low-sigma convergence), RAW + Turbo base models and a 4x upscale pass, with one preset number controlling how far the edit is allowed to move.
Graph (as saved)
89 nodes / 149 links / 38 groups.
<ul>
<li>Depth chain: three <code>Krea2ControlImageEncode</code> stages each feeding <code>Krea2ControlApply</code> and a <code>Krea2ControlLoRALoader</code> on <code>depth-control-lora.safetensors</code> at 1.00 / 0.85 / 0.65 (preset-driven).</li>
<li>Three more <code>LoraLoaderModelOnly</code> slots on the same file are saved at strength 0 - disabled slots, not active LoRAs.</li>
<li>Models: <code>krea2_raw.safetensors</code> high-noise stage, <code>krea2_turbo_fp8.safetensors</code> mid/low. CLIP <code>qwen3vl_4b_fp8_scaled.safetensors</code> (type <code>krea2</code>), VAE <code>qwen_image_HDR_vae_fp32_comfy.safetensors</code>.</li>
<li>Single input: one <code>LoadImage</code>; depth via DA3 nodes; <code>ConditioningKrea2Rebalance</code> saved bypassed in this variant.</li>
<li>Samplers: three <code>SamplerCustomAdvanced</code> with one shared <code>SamplerEulerAncestral</code> (1, 1); shared RAW/Turbo schedule 36 steps denoise 0.92; Turbo upscale refine 10 steps denoise 0.35; active intermediate KSampler 10 steps, CFG 1, euler_ancestral, simple, denoise 0.40. Legacy KSamplers saved bypassed (8j0.90, 10j0.30).</li>
<li>Upscale: 4x-UltraSharp + <code>ImageUpscaleWithModel</code>; two <code>LatentUpscale</code> nodes. SaveImage prefix <code>Krea2_单图洗图_2.4_RAWTurbo</code>.</li>
<li>Preset dial: 0 = tightly controlled / faithful, 1 = balanced (default), 2 = diverging high-noise repaint; drives three denoise values, Split checkpoints, RAW CFG and the depth-LoRA strength trio together (preset 2 sets depth LoRA to 0).</li>
</ul>
How to use
Load one image, pick the preset, and write or extract the prompt. The graph's CLIP node accepts the final prompt from the JSON field; the LLM/JSON caption branch is shared with the batch variants. Leave the depth-LoRA trio at preset values unless you deliberately override it.
Limits and not-verified items
No inference was run for this listing: no speed, VRAM, fidelity or upscale-quality claim is made. The saved nodes and notes are the author's configuration, not measured results. Weights, LoRA, depth model and input media are not bundled; check their licenses separately. The cover image is a promotional still, not a frame produced by this graph.
Tutorial: <a href="https://www.youtube.com/watch?v=aCEuir7zy-w" rel="ugc">YouTube</a> · <a href="https://www.bilibili.com/video/BV1GBgi6JEwL" rel="ugc">Bilibili</a>. Workflows supplied through AI-KSK; verify workflow, node, model, LoRA, input-image and cover rights separately. No blanket commercial or redistribution license is asserted.

krea2Cleanup26SingleImage_v10.json

v1.0

https://youtu.be/aCEuir7zy-w 
 
This workflow is the single-image restyle version of the Krea 2 RAW + Turbo refinement chain. It is meant for creators who want to take one source image and push it into a cleaner, sharper, more polished result while still respecting the original subject, pose, perspective, lighting direction, and general composition. Compared with a full creative redraw, this setup is more practical for one-off restoration, thumbnail cleanup, material refinement, detail recovery, and controlled style enhancement. 
 
The inspected graph uses `krea2_raw.safetensors` for the first high-noise shaping stage and `krea2_turbo_fp8.safetensors` for the later Turbo refinement stages. The text encoder is `qwen3vl_4b_fp8_scaled.safetensors`, loaded in Krea2 mode, and the VAE is `qwen_image_HDR_vae_fp32_comfy.safetensors`. The active control route uses `depth-control-lora.safetensors` across three structural stages, with visible strengths at 1.00, 0.85, and 0.65. This makes the workflow especially useful when the source image already has a strong layout that should not be lost. 
 
The workflow contains a manual JSON prompt field for controlled overrides. The main path includes a shared 36-step RAW/Turbo sigma schedule, RAW high-noise shaping, Turbo intermediate refinement, Turbo low-sigma convergence, latent upscales to 1024 x 1024, and a 10-step Turbo upscale-refine pass. A 4x-UltraSharp image upscaler is present on the active output path. Several older sampler or scale nodes remain in the file as bypassed or disconnected references, so they are not advertised as active behavior. 
 
Main features: 
 
- Single-image Krea 2 restyle workflow 
- Krea 2 Raw high-noise structure rebuild 
- Krea 2 Turbo FP8 low-noise refinement 
- Qwen3VL 4B FP8 Krea2 text encoder 
- Qwen Image HDR VAE decode path 
- Three-stage depth-control chain 
- Depth ControlNet LoRA strengths 1.00, 0.85, and 0.65 
- Manual JSON prompt override 
- 1024 x 1024 latent refinement target 
- Active 4x-UltraSharp upscale route 
- Best for one-image cleanup, sharpening, and controlled restyle 
 
Suggested workflow: 
 
Load one source image and begin with a conservative prompt that says exactly what should be preserved. Treat the original image as the guide for identity, pose, framing, lighting, and color. If the result changes the subject too much, simplify the prompt and keep the preservation language. If the result looks too close to the original, add clearer visual goals such as cleaner edges, richer materials, better surface texture, or a stronger art direction. 
 
Related resources: 
 
Krea 2 Model Collection (Quark): https://pan.quark.cn/s/07bdc81784ce 
Krea 2 Raw: https://huggingface.co/krea/Krea-2-Raw 
Krea 2 Turbo: https://huggingface.co/krea/Krea-2-Turbo 
Krea 2 Depth ControlNet LoRA: https://huggingface.co/Patil/Krea-2-depth-controlnet 
 
RunningHub Workflow 
 
Try the workflow online right now - no installation required. 
Workflow: https://www.runninghub.ai/post/2080681456040501249?inviteCode=rh-v1111 
 
If the results meet your expectations, you can later deploy it locally for customization. 
 
Fan Benefits: Register to get 1000 points + daily login 100 points - enjoy 4090 performance and 48 GB super power. 
 
Bilibili Updates (Mainland China &amp; Asia-Pacific) 
 
If you are in the Asia-Pacific region, you can watch the video below to see the workflow demonstration and creative breakdown. 
Bilibili Video: https://www.bilibili.com/video/BV1GBgi6JEwL/ 
 
Support Me on Ko-fi 
 
If you find my content helpful and want to support future creations, you can buy me a coffee. 
Every bit of support helps me keep creating. 
Ko-fi: https://ko-fi.com/aiksk 
 
Business Contact 
 
For collaboration or inquiries, please contact aiksk95 on WeChat. 
 
打开下方链接即可在线体验，无需安装。 
工作流：https://www.runninghub.ai/post/2080681456040501249?inviteCode=rh-v1111 
 
如果觉得效果理想，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，可体验 4090 / 48G 级别算力。 
 
Bilibili 更新（中国大陆及亚太地区） 
 
如果你在中国大陆或亚太地区，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/BV1GBgi6JEwL/ 
 
本期 Krea 2 相关模型资源： 
Quark: https://pan.quark.cn/s/07bdc81784ce 
Krea 2 Raw: https://huggingface.co/krea/Krea-2-Raw 
Krea 2 Turbo: https://huggingface.co/krea/Krea-2-Turbo 
Krea 2 Depth ControlNet LoRA: https://huggingface.co/Patil/Krea-2-depth-controlnet 
 
这些资源主要面向本地用户，方便进行创作与学习。

Files with SHA256: 7f73d7298f8cbb669c05ece04985793ae5ba02de25636934d4b16df0ef66ebc2