Watch the full video first to understand how this KSK Krea2 original-image redraw and style transformation workflow works in practice. The video shows how one uploaded reference image can be reinterpreted through Krea2, keeping the original character foundation while changing the visual style, setting, atmosphere, materials, and final image quality.

This ComfyUI workflow is designed for Krea2 image-to-image redraw and style transformation. Unlike a pure text-to-image workflow, this version starts from an existing image. The uploaded image is used as the visual base, then the prompt guides Krea2 to rebuild the character into a new polished artwork. This makes the workflow useful for character redesign, style upgrade, visual transformation, AI influencer redesign, fantasy character conversion, poster-style image refinement, and reference-based artwork creation.

The workflow uses krea2_turbo_bf16.safetensors as the main generation model. The text encoder route uses qwen3vl_4b_fp8_scaled.safetensors with the Krea2 CLIP type, while the VAE route uses qwen_image_vae.safetensors. This gives the workflow a compact Krea2 Turbo image-to-image structure that is fast enough for practical iteration.

The input route is the key part of this workflow. A reference image is loaded through LoadImage, then processed by image_scale_pixel_v2 before being encoded into latent space through VAEEncode. This latent becomes the starting point for the KSampler. Because the workflow is image-to-image based, the result can preserve the broad character identity, pose, composition, and silhouette while still allowing a new style direction.

The prompt is written as a transformation instruction. It asks the model to preserve the same character identity and main visual traits while upgrading the artwork into a polished vertical fantasy character image. This is the correct usage pattern for this workflow: first tell the model what must remain, then tell it what should change, and finally define the new environment, lighting, materials, rendering style, and quality target.

The workflow uses ConditioningKrea2Rebalance to strengthen the positive conditioning. The Rebalance setup uses a custom 12-layer weight structure and a multiplier of 2.5. This helps the transformation prompt become more effective without pushing the image too aggressively. Compared with the stronger multiplier 4.0 workflows, this version is more controlled and better suited for redraw tasks where the original image should still remain recognizable.

The sampling setup uses KSampler with 8 steps, CFG 1, euler sampler, simple scheduler, fixed seed, and denoise 0.5. The denoise value is important here. It gives the model enough freedom to redraw style, materials, and scene details, while still keeping a relationship with the source image. Higher denoise would create a stronger transformation but may weaken identity preservation; lower denoise would keep the image closer but reduce style change.

The output route is straightforward: the sampled latent is decoded through VAEDecode, previewed through PreviewImage, and saved through SaveImage with a Krea2 image-to-image prefix.

Main features:

• Krea2 image-to-image redraw workflow

• Original image based style transformation

• Reference image identity preservation

• Character redesign and visual upgrade

• krea2_turbo_bf16.safetensors support

• qwen3vl_4b_fp8_scaled.safetensors text encoder

• qwen_image_vae.safetensors VAE

• LoadImage reference input

• image_scale_pixel_v2 preprocessing

• VAEEncode source image to latent

• ConditioningKrea2Rebalance enhancement

• Custom 12-layer Rebalance weights

• Multiplier 2.5 balanced control

• 8-step fast generation

• CFG 1 setup

• euler sampler

• simple scheduler

• denoise 0.5 image-to-image strength

• VAEDecode final image decoding

• PreviewImage and SaveImage output

• Suitable for character redesign, style conversion, and fantasy artwork transformation

Suggested workflow:

Start with a clean reference image. The character should be readable, with a clear face, pose, outfit, and silhouette. Then write the prompt in three parts: what to preserve, what to transform, and what final style to create. Keep identity-preservation words near the beginning of the prompt, then describe the new clothing, setting, lighting, material, mood, and rendering style. Use denoise 0.5 as the balanced starting point. Raise it only when a stronger redesign is needed. Lower it when the source identity must stay closer. This workflow is best used for turning an existing character image into a more polished Krea2 fantasy, sci-fi, anime-realistic, or cinematic artwork.

⚙️ RunningHub Workflow

Try the workflow online right now — no installation required.
👉 Workflow: https://www.runninghub.ai/post/2072199218311286785?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 & Asia-Pacific)

If you’re 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/BV1FWT76KENn/

☕ 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 — just like a spark that can ignite a blazing flame.
👉 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/2072199218311286785?inviteCode=rh-v1111
如果觉得效果理想，你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 48 G 超级性能！

📺 Bilibili 更新（中国大陆及南亚太地区）

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/BV1FWT76KENn/

我会在 夸克网盘 持续更新模型资源：
👉 https://pan.quark.cn/s/20c6f6f8d87b
这些资源主要面向本地用户，方便进行创作与学习。