<div data-youtube-video="https://youtu.be/CU1UKxjWBRo"><iframe src="https://www.youtube.com/embed/CU1UKxjWBRo" width="853" height="480" allowfullscreen></iframe></div> 
RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2079971231356506114?inviteCode=rh-v1111
<h2>Krea 2 Identity Edit v1.2 - Two-Image Reference Editor</h2>
Purpose: the two-image variant of the Krea 2 Identity Edit v1.2 graph. Image A is the scene, image B is the reference (a person, an outfit, a product), and the instruction tells the model how to bring B into A while preserving the identity and the rest of the scene.
<h2>Graph (as saved)</h2>
22 nodes / 31 links.
<ul>
<li>UNETLoader: <code>krea2_turbo_fp8_scaled.safetensors</code>; LoRA: <code>krea2_identity_edit_v1_2_fp16.safetensors</code> at strength 1.</li>
<li>CLIPLoader: <code>qwen3vl_4b_fp8_scaled.safetensors</code>, type <code>krea2</code>; VAE: <code>qwen_image_HDR_vae_fp32_comfy.safetensors</code>.</li>
<li>KSampler: 10 steps, CFG 1, <code>euler</code>, <code>simple</code>, denoise 1.</li>
<li>Krea2EditModelPatch: ref_boost 3, two-image mode, fit.</li>
<li>Two LoadImage inputs are active (image A = jpg, image B = webp in the saved example); the example instruction transfers a complete sporty outfit from image B onto the woman in image A, and names facial identity, hairstyle and the original scene as things to preserve.</li>
<li>Two PreviewImage nodes plus SaveImage prefix <code>ComfyUI/Kera2Edit</code>.</li>
</ul>
<h2>How to use</h2>
Put the scene in image A and the reference in image B, then write the transfer in plain English. For people-in-scene edits, name the subject ("the woman in Image A") and list what must stay unchanged. Two-person inputs are better kept at or below about 1.5 MP per the workflow notes; single-subject edits can go higher.
<h2>Limits and not-verified items</h2>
No inference was run for this listing: no speed, VRAM, identity-transfer quality or compositing claim is made. The saved images, prompts and previews are the author's demo content, not evidence of an output. Two-image behaviour depends on the exact Krea2EditModelPatch implementation; verify it on your own material. Weights, LoRA 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=CU1UKxjWBRo" rel="ugc">YouTube</a> · <a href="https://www.bilibili.com/video/BV1bJgy6eEei" rel="ugc">Bilibili</a>. Workflows are supplied through AI-KSK; verify workflow, node, model, LoRA, source-image and cover rights separately. No blanket commercial or redistribution license is asserted.

Krea 2 Identity Edit v1.2 - Two-Image Reference Editor

<div data-youtube-video="https://youtu.be/CU1UKxjWBRo"></div> 
RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2079971231356506114?inviteCode=rh-v1111
Krea 2 Identity Edit v1.2 - Two-Image Reference Editor
Purpose: the two-image variant of the Krea 2 Identity Edit v1.2 graph. Image A is the scene, image B is the reference (a person, an outfit, a product), and the instruction tells the model how to bring B into A while preserving the identity and the rest of the scene.
Graph (as saved)
22 nodes / 31 links.
<ul>
<li>UNETLoader: <code>krea2_turbo_fp8_scaled.safetensors</code>; LoRA: <code>krea2_identity_edit_v1_2_fp16.safetensors</code> at strength 1.</li>
<li>CLIPLoader: <code>qwen3vl_4b_fp8_scaled.safetensors</code>, type <code>krea2</code>; VAE: <code>qwen_image_HDR_vae_fp32_comfy.safetensors</code>.</li>
<li>KSampler: 10 steps, CFG 1, <code>euler</code>, <code>simple</code>, denoise 1.</li>
<li>Krea2EditModelPatch: ref_boost 3, two-image mode, fit.</li>
<li>Two LoadImage inputs are active (image A = jpg, image B = webp in the saved example); the example instruction transfers a complete sporty outfit from image B onto the woman in image A, and names facial identity, hairstyle and the original scene as things to preserve.</li>
<li>Two PreviewImage nodes plus SaveImage prefix <code>ComfyUI/Kera2Edit</code>.</li>
</ul>
How to use
Put the scene in image A and the reference in image B, then write the transfer in plain English. For people-in-scene edits, name the subject ("the woman in Image A") and list what must stay unchanged. Two-person inputs are better kept at or below about 1.5 MP per the workflow notes; single-subject edits can go higher.
Limits and not-verified items
No inference was run for this listing: no speed, VRAM, identity-transfer quality or compositing claim is made. The saved images, prompts and previews are the author's demo content, not evidence of an output. Two-image behaviour depends on the exact Krea2EditModelPatch implementation; verify it on your own material. Weights, LoRA 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=CU1UKxjWBRo" rel="ugc">YouTube</a> · <a href="https://www.bilibili.com/video/BV1bJgy6eEei" rel="ugc">Bilibili</a>. Workflows are supplied through AI-KSK; verify workflow, node, model, LoRA, source-image and cover rights separately. No blanket commercial or redistribution license is asserted.

krea2IdentityEditV12Two_v10.json

v1.0

https://youtu.be/CU1UKxjWBRo 
 
This workflow is the two-image editing setup for Krea 2 Identity Edit v1.2. It is built for edits where one reference image is not enough: replacing a person with another reference, transferring an outfit, combining a subject with a second visual cue, or building a scene from two separate inputs. The workflow gives creators a direct way to test multi-reference editing without manually rebuilding a complicated graph every time. 
 
The inspected graph uses `krea2_turbo_fp8_scaled.safetensors` as the active UNET, `qwen3vl_4b_fp8_scaled.safetensors` as the Krea2 CLIP/text encoder, and `qwen_image_HDR_vae_fp32_comfy.safetensors` as the VAE. The active LoRA is `krea2_identity_edit_v1_2_fp16.safetensors`, enabled at strength 1.0. The main output path contains a 1024 x 1024 latent setup and a 10-step KSampler with CFG 1, euler/simple sampling, and denoise 1. The graph also includes preview outputs for checking each reference side before the final SaveImage output. A separate ImageScaleBy node exists in the file, but it is bypassed and not part of the active output path, so this description does not present it as an enabled upscale stage. 
 
This version is most useful when image A provides the base scene or target composition and image B provides the identity, face, outfit, object, or visual feature you want to transfer. Keep the prompt explicit about which image supplies which part of the result. Short prompts often work well for replacement tasks, but clear role assignment is important: for example, replace the character in image A with the person from image B, or apply the outfit from image B to the subject in image A. 
 
Main features: 
 
- Two-image Krea 2 edit workflow 
- Multi-reference input logic 
- Krea 2 Turbo FP8 active UNET path 
- Krea 2 Identity Edit v1.2 LoRA enabled at strength 1.0 
- Qwen3VL 4B FP8 Krea2 text encoder route 
- Qwen Image HDR VAE decode path 
- 1024 x 1024 latent setup 
- 10-step euler/simple KSampler 
- CFG 1 and denoise 1 edit pass 
- Multiple preview outputs before final save 
- Best for replacement, outfit transfer, and two-reference scene edits 
 
Suggested workflow: 
 
Use image A as the base frame and image B as the reference you want to borrow from. In the prompt, clearly state the role of each image. If the result mixes the two references incorrectly, simplify the prompt and separate the instruction into base image, transferred feature, and final style. If identity matching is more important than style, keep the style language minimal and focus the prompt on the subject replacement or face/outfit transfer. 
 
RunningHub Workflow 
 
Try the workflow online right now - no installation required. 
Workflow: https://www.runninghub.ai/post/2079971231356506114?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/BV1bJgy6eEei/ 
 
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/2079971231356506114?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，畅玩 4090 体验 48G 超级性能。 
 
Bilibili 更新（中国大陆及亚太地区） 
 
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/BV1bJgy6eEei/ 
 
本期 Krea 2 相关模型资源： 
https://pan.quark.cn/s/07bdc81784ce?pwd=FpDf#/list/share 
 
这些资源主要面向本地用户，方便进行创作与学习。

Files with SHA256: 484d759a0bbb5e4708ccedb2f26b014590378978c9b9c9c2ffe7f2e067a9480f