阿里通义实验室推出的全球最强开源图像生成模型Z-Image，果然名不虚传，中文提示词的理解能力超强，而且还能准确无误的在图片中生成中文文字。而且速度非常快！
This is a workflow for Z-Image,which has amazing accuracy for Chinese&English prompts,and it can precisely display the words(even mandarin charactors)in your image.Further more,it works very fast!
About IMG2IMG version:
You may switch from the 2 different “latent” nodes to switch easily from txt2img to img2img.Remember to adjust "denoising" at the same time.
通过切换“latent”节点，可以轻松的从文生图切换到图生图。注意同时切换“降噪”数值。
Description
FAQ
Comments (12)
真的太好用了！！！
Is it possible to add more LORAs to the workflow?
I think yes,you may try to add nodes for loras and use strings to connect them in a row)
I saw such components in other workflows)
@htshc204736 Thanks! It worked, however Im still having a bad time testing the 3 LORA's weights hehe, again I appreciate the sharing work!!
Damn, new to this and trying to get IMG2IMg to work Spent hours trying to figure it out. I always get the outcome looking like someone else not my original photo. If I put Denoise high I get NSFW but again someone else.
Have you switched the connnetion line which links "latent"?
@htshc204736 Im a beginner, I'm not sure what you mean, sorry I thought this workflow is ready and there are a few latent button so im not sure what needs to be swapped out.
@smellvir151 You need to switch button of the purple line from "空Latent SD3” to "VAE编码”（VAE coding）which links to “K采样器”（K sampler）when you switch from TXT2IMG to IMG2IMG
Image posted by htshc204736
Maybe this pic helps you better understand
Sorry because I'm new, I thought that img2img will include face swap. I misunderstood the model and work flow. My aim was to upload an image and use the image likeness (Face/Body) to make another photo.
If only general consistency of clothing and background setting is required,then the fairly detailed description prompts can be achieved using text-to-image editing. However, if precise consistency of facial features is required, more precise methods are likely needed.