HiDream E1.1 Image Editing by Prompt
Load an image, type what you want changed, and HiDream E1.1 produces the edited version. A native ComfyUI implementation with no custom nodes.
- L40S
- Image editing
- Free template
How it works
The workflow is built entirely from core ComfyUI nodes. UNETLoader loads hidream_e1_1_bf16.safetensors, and QuadrupleCLIPLoader loads the four text encoders HiDream needs: clip_l_hidream, clip_g_hidream, t5xxl_fp8_e4m3fn_scaled and llama_3.1_8b_instruct_fp8_scaled. The VAE is ae.safetensors.
Your image is loaded with LoadImage and resized with ImageScaleToTotalPixels. InstructPixToPixConditioning combines the positive prompt, the negative prompt, the VAE and the image into conditioning for editing. RescaleCFG adjusts the model output, and DualCFGGuider guides the sampling.
SamplerCustomAdvanced runs with RandomNoise, KSamplerSelect and BasicScheduler, then VAEDecode produces the final image. The default negative prompt is simply low quality, blurry, distorted.
Tips
- Write the prompt as an instruction describing the change, for example what to recolor or add.
- The workflow includes links to the model files in its notes, so check that all files are in place if a node reports an error.
- Because
ImageScaleToTotalPixelssets the working size, smaller values run faster at the cost of detail. - Keep the edit focused on one change at a time for more predictable results.
- The model files are large because HiDream uses four text encoders, so the first run on a fresh machine takes longer than later runs.
What you can use it for
- Change colors or objects with a plain instruction
- Style changes to a photo
- Quick edits without drawing a mask
- Compare HiDream editing with other editors
Run it in three steps
Open the template
Click Run on ComfyDeploy. The workflow, its custom nodes and models are set up for you on a cloud machine.
Add your inputs
Provide image to edit and edit instruction prompt in the playground or the ComfyUI editor.
Run, share or deploy
Run it on an L40S, share it with your team as a playground, or deploy it as an API endpoint.
Frequently asked questions
What GPU does this HiDream E1.1 workflow need?
It is set to run on an L40S GPU on ComfyDeploy.
Which models does it load?
It uses hidream_e1_1_bf16 as the diffusion model, with clip_l, clip_g, t5xxl and llama_3.1_8b_instruct text encoders, and ae.safetensors as the VAE.
How do I run it as an API?
Deploy it on ComfyDeploy and expose the image and prompt inputs. The workflow has no custom nodes, so setup is straightforward.
