Watch the full video first if you want to understand how this Boogu-Image 0.1 single-image editing workflow works in practice. The video shows how one uploaded image can be edited through a direct instruction prompt, while the workflow keeps the original image structure, subject identity, pose, camera angle, and overall composition more stable.

This ComfyUI workflow is designed for Boogu-Image 0.1 single-image editing. Its main purpose is to modify an existing image by combining the source image with a natural-language editing instruction. Instead of generating an image completely from scratch, this workflow uses the uploaded image as the visual foundation and applies targeted changes through the Boogu edit model.

The workflow is built around boogu_image_edit_fp8_scaled.safetensors as the main image editing model. The text and image understanding route uses qwen3vl_8b_fp8_scaled.safetensors loaded as a Boogu CLIP encoder. The VAE route uses ae.safetensors for final decoding. This makes the workflow simple, direct, and suitable for fast image editing tests.

The image input section uses a standard LoadImage node. The user uploads one source image, and that image is sent into TextEncodeBooguEdit. This node is the core editing interface. It receives the source image, the Boogu CLIP encoder, the VAE, the prompt, and the negative prompt, then converts the editing instruction into positive and negative conditioning for the sampler.

The prompt is separated into an independent Text node before entering TextEncodeBooguEdit. This is useful because creators can expose the prompt field more cleanly in RunningHub or a simplified application interface. Users only need to upload an image and describe what they want to change. For example, the workflow can replace clothing, adjust style, change visual elements, or refine specific image details while asking the model to preserve the face, hairstyle, pose, and camera angle.

The generation canvas is created through EmptyLatentImage at 1024×1024 with batch size 1. The model route uses ModelSamplingAuraFlow with shift set to 3.16. The sampler is dpmpp_2m, and the BasicScheduler uses the simple scheduler with 35 steps and denoise set to 1. The SamplerCustom node uses CFG 3.5, which gives the prompt enough influence while keeping the edit route controlled.

After sampling, VAEDecode converts the latent output into the final image. PreviewImage is used for immediate visual checking, and SaveImage exports the edited result. The overall structure is clean and easy to reuse because it avoids unnecessary repair chains, extra ControlNet branches, or complex post-processing modules.

This workflow is suitable for single-image editing, clothing replacement, character detail adjustment, object modification, visual refinement, style change, product image edits, AI portrait editing, and controlled creative retouching. It is especially useful when you want a direct “upload image + write instruction + get edited image” experience.

Main features:

• Boogu-Image 0.1 single-image editing workflow

• One uploaded image as the editing source

• boogu_image_edit_fp8_scaled.safetensors main model

• qwen3vl_8b_fp8_scaled.safetensors Boogu encoder

• ae.safetensors VAE decoding

• TextEncodeBooguEdit instruction-based editing

• Independent Text node for editable prompt input

• 1024×1024 latent canvas

• ModelSamplingAuraFlow with shift 3.16

• dpmpp_2m sampler

• Simple scheduler with 35 steps

• SamplerCustom CFG 3.5

• PreviewImage for checking results

• SaveImage for final export

• Clean and reusable image-editing structure

Suggested workflow:

Upload one clear source image first. Then write a direct editing instruction that describes what should change and what should remain unchanged. For best results, separate the edit target from the preservation rule. For example, describe the new clothing, object, color, background, or style first, then clearly state that the face, hairstyle, pose, composition, lighting, and camera angle should remain basically unchanged. Avoid giving too many conflicting changes in one prompt. If the edit is too weak, make the instruction more specific. If the edit changes the identity too much, strengthen the preservation wording and simplify the requested modification.

⚙️ RunningHub Workflow

Try the workflow online right now — no installation required.
👉 Workflow: https://www.runninghub.ai/post/2067856537388933122?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/BV17ujH6eEZk/

☕ 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/2067856537388933122?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/BV17ujH6eEZk/

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。