V1.1
Updated the dataset generation prompts to improve variation across facial expressions, lighting, camera angles, poses, and perspectives instead of keeping the face too close to the original reference expression.
I also simplified the reference-image setup:
Select A for 1 reference image
Select B for 2 reference images
On my RTX 5080, generating 20 images with 1 reference image of around 1 MB took roughly 10 minutes.
Overall, V1.1 should produce a more diverse and useful dataset for LoRA training, with stronger variation in expressions, lighting, poses, and viewpoints while still preserving the subject's identity.
Change max rows to 25 if u want 25 images and you can always change the prompts send the prompt used to chatgpt or somewhere u like to get full body shots or different styles
-------------------------------------
Qwen-Image 2.1 Dataset Generation Workflow
This is an updated/adapted version of an existing dataset-generation workflow originally created by acekiube. Full credit for the original workflow structure and idea goes to them. I did not create the workflow from scratch.
I modified the workflow to work with Qwen-Image 2.1, updated the model/encoder nodes, generation settings, resolution handling, and prompts for identity-focused dataset generation.
The goal is to take a reference character/person and automatically generate a set of useful training images with different viewpoints, expressions, camera angles, and lighting while keeping the identity as consistent as possible.
I’ve personally had better identity consistency around CFG 3, especially for side/profile views. The workflow is currently set to CFG 3, 25 steps, res_multistep, and a 3:4 1MP output.
I’m also experimenting with giving Qwen additional reference views, such as a side-profile image, since this seems to help significantly when generating difficult angles.
If you got better prompt or settings please let me know
Models used:
Qwen2.1 int8 convrot:
https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/diffusion_models/qwen_image_2.1_int8_convrot.safetensors
Text Encoder:
https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/text_encoders/qwen3vl_8b_int8_convrot.safetensors
Vae:
https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/vae/qwen_image_2.1_vae_bf16.safetensors
Description
FAQ
Comments (13)
thanks for this, hopefully this will come in handy for LORA training
First model, after a lot of Klein2 & Krea 2 dataset/datasheet workflows, where I have likeness and not random faces for the references image. Did not try with sideprofile yet, just bypassed this node on my first run. I honestly thought my subject simply wasn't datasheet-able for whatever reason. Awesome.
EDIT: Ok, with 2 reference images I am at 1 hour after the 5th xD
You probably put very high quality ref images as the higher quality the longer. Thanks for the advice ill fix it
@Jolanoff I think especially the second one had a very high quality because I upscaled it some time ago. Thanks for the hint and for fixing it.
@Jolanoff Report back for v1.1
- For using 2 References (B Mode): There are multiple generated images, which consist of up to 14 sub images. Dunno how to describe, but they are not "standalone" :D
- I ran into the trap again with image sizes. I added 2 Resize Image/Mask nodes, scale total pixels, 1.00 megapixels, area method
- KJ's output_folder only takes absolute paths or "output" for comfyuis output folder. subfolders etc need to be added to filename_prefix
- Make caption files optional please. I disconnected the path for now.
@daze45230 i tried image scale down to size node and put it at 1024. the results were good, but you get some quality loss too, you could try that for the large images,..
im gonna check the B problem with multiple images and make captions optional.
I will see what i can do abt the kj output issue
@daze45230 update : i removed the caption files totally cuz i think they arent necessary as every lora creates that for you anyways.. and fixed the issue with 2 images. and added some labels
Hi
It's not working for me.
I waited about 7 minutes after pressing the "Run" button, but it didn't produce any output.
Did you put an output folder directory ?
Same here, waited for 10 minutes, doesn't work. Output directory set, no changes.
@sarahbonestone if u change the max rows under CR prompt list to 2 or 1. and show ur log after. it would be helpful
@Jolanoff no output either
comfyui alredy has the most simple workflow guys please u dont need this