288 Poses, 1,407 Outfits: Two Free Libraries Take Apart the Reference Images for AI Image Generation

Two free reference libraries were recently recommended together. One handles poses, and the other handles clothing. For people making AI images, this solves the very time-consuming task of finding reference images.

First, the pose library. The Spanish-language tweet's author, roman, says it has 288 poses and more than 1,439 reference images, covering angles such as front, side, top-down, and bottom-up, as well as full-body poses and different actions. The usage is straightforward: take the pose image you want as a reference and feed it into an AI tool's image-reference or pose-control feature so the model knows how the person should stand and move. Link: https://ai-pose-gallery.pages.dev/

The video demonstrates the combined usage. On the right side of the chat interface are three images: one of a woman, one of black clothing, and one of a human-body line-art pose. The Japanese instruction below is “1枚目の女性を2枚目の服装、3枚目のポーズに変えて”, meaning: change the woman in the first image into the clothing in the second image and pose her according to the third image. This workflow splits pose and clothing into two variables; changing one does not require finding a whole new set of references for the other.

Next, the clothing library. The page shows 1,407 model-free clothing images, categorized by spring, summer, autumn, winter, underwear, party, swimwear, and sexy. The image ratio is 2:3, and they contain only clothes, with no faces, body shapes, or backgrounds. You can filter by category and also search by ID or terms such as denim, lace, or white. Downloading the original PNG is more reliable for checking details. Link: https://ai-costume-gallery.pages.dev/

The usage notes on the clothing library page are fairly clear: free to use; commercial use, editing, and referencing in creating works are OK; the images themselves may not be redistributed, reposted, or resold; bulk downloading is prohibited, including automated collection with tools; no attribution required; the images are fictional clothing generated by t2i, not photos of real products; similarity checks against existing brands were done before release, but there is no guarantee of non-infringement of third-party rights; use at your own risk; do not use for real people, minors, or purposes that violate public order and good morals.

The combination method is not complicated: first choose a pose, then choose clothing, and give the two reference images separately to the AI tool. Its value is not in the number of images but in turning reference images into searchable assets. To find a “low-angle, full-body, side-standing pose,” you don't have to scroll through an album for ages; to find “spring, lace, white, mini skirt,” you can search directly. However, the clothing library has no people, so faces, body types, and backgrounds still need separate references or instructions. If you expect one link to solve character consistency, you will most likely be disappointed. The time saved by free resources may end up being spent filtering images.

Some users said in replies that putting the two libraries together is a plus; others found the pose library especially useful. Original tweet link: https://x.com/Nozelcode/status/2116917483365883921. The clothing library page lists the creator as ギア. Save it first and use it as needed.