Full tutorial of how to train and use published :
Trained with SECourses Musubi Trainer Premium app : https://www.patreon.com/posts/secourses-musubi-137551634
Trained model is Qwen Image Edit Plus 2509. Can be used without reference or input images just as text to image.
Description
Full tutorial of how to train and use published : https://youtu.be/DPX3eBTuO_Y
Trained with SECourses Musubi Trainer Premium app : https://www.patreon.com/posts/secourses-musubi-137551634
Trained model is Qwen Image Edit Plus 2509. Can be used without reference or input images just as text to image.
FAQ
Comments (18)
The results are amazing. How much of a difference was there between a lora train, and a whole checkpoint train? I can't argue with the results, but were the lora results that bad?
Also, you say this: "Trained model is Qwen Image Edit Plus 2509. Can be used without reference or input images just as text to image."
Does this mean it cannot be used with reference? Are the results not as good? Or does it not try? If it works, what would the trigger be? "Convert image to ohwx"?
Anyway, I hope it's clear that I'm very impressed, which is why I have so many questions. I don't mean to sound like I'm poking holes. It's so good that it makes me wonder if I should change my approach, so I want to know more about it.
Hey @Jellai thanks for comment. LoRA quality almost same. I am preparing a tutorial so you will see. You can use with reference images as well. But it is not mandatory. For example on myself trained model I am able to command as replace head with ohwx man and it works
@SECourses Looking forward to it. Also, I wonder what I said wrong that someone gave me a thumbs down. Oh well.
@Jellai ye haters are everywhere :D
How is this style at doing women? I see a ton of gruff men.
it should be same i didn't do any examples
Am I missing something? All this does is generate a pixelated garbled image, no matter what settings I use. Do you have a workflow you can share? None of your example images will import the workflow.
Hello. I just published the full tutorial and shown how to generate amazing quality images as well with SwarmUI preset : https://youtu.be/DPX3eBTuO_Y
@SECourses Everything about the lesson is paid. It's so easy to share the workflow, otherwise it doesn't work. Why post something that doesn't work?
@Eugenich I managed to get it working. You need to generate an script with the llm to fix the metadata issues. The model here is broken due to the models metadata. You need to get an llm to generate a script that goes through the model and fixes the metadata issues. I have a script somewhere but I basically just got it to figure out the issue and generate a script to fix the model. If I can find my script i'll post it but if you show some of the error messages and debug logs from swarmui to gemini or whatever and give it a hint that there's a metadata issue with the model then it should be able to make the right script to fix it.
@DankOwl interesting the metadata kohya generates works directly in swarmui and comfyui for me. where do you use?
@SECourses I messaged you on patreon a while back with this issue and the solution. I just wanted to give people my solution on this post just in case people thought it was broken still which it seems like a lot of people were reporting my exact issue. Yes it was in swarm UI.
Full tutorial of how to train and use published : https://youtu.be/DPX3eBTuO_Y
not work
i would do this with chat, but if you take me up on this offer, i want the starting point of our success to be public. Hire me and finally reach a level of success both of us deserve. You are missing one thing all the time, the work of a fine artist, it is cute to see you but it stop selling products a long time ago (plus I believe you would be one to want to help a fine artist survive, we are the reason you can make the images in first place). This field is tough and everyone demands the best, together we would deliver that. I need a job, you need an artist, give me a job, and we will shake the ai community, those that pass you up will take a second look, and i finally have time to show what a fine artist can do with the tools built from our minds. : )
i think your ai bot replied to wrong post
@SECourses nope, i am a fine artist, that is just how i am, checking to see if your interested to help yourself and support the fine arts, not many of us here, the source of all ai, we fine artist, are the source, check my profile