Z-image T2I
PixelArt
Welcome to try it out, and have fun everyone！Tell me what you think if you want!
If you'd like to quickly experience the model in the cloud or refer to my workflow, check it out here. The prompts are included as well. There is more detailed information about the model if you are interested.：
T2I quick experience:
https://www.runninghub.ai/ai-detail-new/2017875450126667777/?inviteCode=rh-v1395
T2I workflow:
https://www.runninghub.ai/post/2017582166745554946/?inviteCode=rh-v1395
Applications
This model can only be used for T2I.
Trigger word
Not required
Workflow
Since it's a base model, I think a single sampling pass should suffice. After all, it's really too slow.
Recommended Weight: 1.1 ～ 1.3
Model: Base version
Steps: 50
CFG: 3.0
Tips
The training of this model is based on a fascinating open-source project. Feel free to check out the link if you're interested:https://huggingface.co/DiffSynth-Studio/Z-Image-i2L
The workflow retains the I2L (Image-to-Lora) section. Feel free to experiment and train your own preferred styles.
Description
Based on the open-source project: Z-Image-i2L
Trained on 10 images
FAQ
Comments (4)
Just 10 images? Can I ask what hw you used and how long it took?
ComfyUI on the cloud with a 48GB VRAM setup, and the process takes about 90 seconds.
@Crazy0range I think they meant how long did the training took. There no chance it takes only 90 seconds is there? If so would you mind referring me to a learning source? I am interested in training my own pixel art lora for z-image base, thanks!
@Derpervirens Sorry for the late reply, I've been quite busy recently. The 'training' process does indeed take about 90 seconds because this isn't conventional training in the usual sense. It is based on a fascinating open-source project.
As I mentioned in the TIPS section of the introduction: 'The training of this model is based on a fascinating open-source project. Feel free to check out the link if you're interested: https://huggingface.co/DiffSynth-Studio/Z-Image-i2L
The workflow retains the I2L (Image-to-Lora) section. Feel free to experiment and train your own preferred styles.'