The model is not restricted for commercial use, does not require membership downloads, and does not set encrypted downloads.

However, I don't want my models to be picked up and used directly by others, so I record a tutorial for each model:

https://www.douyin.com/video/7510889573224910099?modeFrom=userPost&secUid=MS4wLjABAAAA7EDFaK4UYRM5m8mmLOn6qPNjFzlwmaP8LUCdKtzwWSs

This is a puzzle lora that achieves character consistency. It can generate 4 to 6 consistent images of characters, clothing, and scenes at once, which can later be enlarged, cropped for individual optimization, or used for face-swapping. All of these can easily result in a portrait photography set. The images above are either direct outputs or direct outputs with enlargement. If there are issues with the hands, you can crop and fix them individually.

—— Usage:

Mainly watch the video above. The corresponding workflow can be downloaded from my homepage. The weight is 0.7, and the size is 1024*1536.

Prompt words are quite important, and the template can be modified as needed: This photo collage features a young Asian woman with fair skin and black hair styled into a loose bun with a hair clip. She is wearing a light yellow long-sleeve off-shoulder top and white shorts, paired with white socks. The collage showcases her in four different poses in a cozy, sunny living room. She lies on a beige carpet, holding a green apple, interacting with a large beige teddy bear, and sitting on the floor. The background includes a wooden coffee table, an old-fashioned black TV, and a bookshelf with books and a small plant. The lighting is soft and warm, creating a relaxed and cozy atmosphere.

An easier way is to let a large language model generate similar prompts for you, and you can use them directly. I haven’t tried the online web UI, so feel free to test it yourself.

Issues with hands are minor and are generally automatically fixed during the enlargement process. The key is to select suitable compositions and correct body postures.

I hope everyone can leave comments and share their results. I will update the next model as soon as possible, striving to make each model valuable and unique.