Page 1 of 1

Transform your portrait into a retro anime style

Posted: Thu Jan 23, 2025 8:43 am
by Rina7RS
At the same time, because the internal generation process has been optimized, the speed of generating pictures is 10%~25% higher than that of webui (the increase varies depending on different graphics cards). When generating large pictures, the video memory will not explode. However, when the picture is too large, it will be fragmented due to the block operation.

This case builds a workflow based on ComfyUI, which can transform the portrait into the style of the comic masters of the golden age of Japanese comics. Case experience link Convert the portrait into a retro Japanese comic style in ComfyUI

Features: Retro Japanese comic style
Large model: Animagine XL V3.1
Lora weight recommendation: 0.65-0.8
Number of steps: 25-35 steps
CFG: 7
Sampling method: DPM++ 2M Karras

2. Preliminary work
In this case, the ComfyUI service running in the notebook jordan phone number data needs to be open to the public network for use, which requires the use of ngrok. Users can go to the ngrok website to register an account for free, and after logging in, select "Verification" in the list on the left, and remember your authtoken, which will be used later.

Precautions
Click the "Run in ModelArts" button in the upper right corner to enter ModelArts CodeLab. You will need to log in to your Huawei Cloud account. If you do not have an account, you will need to register one and complete real-name authentication. Please refer to "How to create a Huawei Cloud account and complete real-name authentication" to complete account registration and real-name authentication. After logging in, wait for a while and you can enter the CodeLab operating environment.