Consistent characters: the four methods, ranked by effort

"How do I get the same character in every image" is the most-asked question in AI art. There are four real methods, and they stack. 1. A fixed character record (zero effort) Write the character once, in a fixed order, and paste it verbatim every time: age, build, skin, face shape, hair colour and style, eye colour, one signature item. Change only the scene words. This alone gets you a family resemblance, not the same person. 2. Reference image (low effort) A reference image steers the new one. IP-Adapter is the standard tool: https://github.com/tencent-ailab/IP-Adapter Hathor Studio's Character Effects and Appear with Hathor use this: upload the character's picture and describe a new scene. Best balance of effort and result. 3. Reference image plus pose or layout control (medium) IP-Adapter for who, an openpose or depth ControlNet for where and how. This is how you get the same character in a specific pose. 4. A trained LoRA (high effort, best result) Train a small add-on model on 15-30 images of the character. Then a trigger word summons them reliably in any scene. See the LoRA training thread on this board and https://wiki.soapbox.community/wiki/LoRA_and_Fine_Tuning What does not work Reusing the seed. A seed gives the same starting noise; change the prompt and you get a different face. The Studio's own method The Studio's Make it yourself page teaches: make the character first, then the things (objects and places, each on its own), then compose the scene — rather than hoping one prompt gets everything right. The Reference Studio at https://hathor.soapbox.community/compose lets you upload several references at once, each tagged as character, object or scene.