Midjourney launches feature to generate consistent characters across multiple AI images from multiple generations

Join leaders In Boston on March 27 For A exclusive night of networking, knowledge, And conversation. Request A invite here.

THE popular AI picture generator service Mid Road has deployed A of It is most often asked features: THE ability has recreate characters regularly through new images.

This has has been A major obstacle For AI picture generators nowadays, by their very nature.

It is because most AI picture generators rely on on "diffusion models," tools similar has Or base on Stability AI Stable Diffusion Open source picture generation algorithm, which work roughly by socket text grasped by A user And trying has piece together A picture pixel by pixel that matches that description, as learned Since similar imagery And text Keywords In their massive (And controversial) training data together of millions of human created images.

For what consistent characters are SO powerful — And elusive — For generative AI imagery

Again, as East THE case with text based big language models (LLM) such as OpenAI ChatGPT Or Cohere new Order, THE issue with all generative AI apps East In their inconsistency of answers: THE AI generates something new For each Single fast between In he, even if THE fast East say again Or a few of THE even key words are used.

V.B. Event

THE AI Impact Tour – Boston

Were excited For THE following stop on THE AI Impact Tour In Boston on March 27. This exclusive, invite only event, In Partnership with Microsoft, will functionality discussions on best practices For data integrity In 2024 And beyond. Space East limit, SO request A invite today.

Request A invite

This East great For generator entire new pieces of content — In THE case of Half-way, pictures. But What if You are storyboard A movie, A novel, A chart novel Or comic book, Or a few other visual AVERAGE Or You to want THE same character Or characters has move through he And appear In different scenes, settings, with different facial care expressions And accessories?

This exact scenario, which East typically necessary For narrative continuity, has has been very difficult has reach with generative AI — SO far. But Mid Road East NOW socket A crack has he, presentation A new label, “–cref” (short For "character reference") that users can add has THE END of their text instructions In THE Mid Road Discord And will to try has match THE characters facial care features, body type, And even clothes Since A URL that THE user pasta In following said tag.

As THE functionality progress And East refined, he could take Mid Road further Since be A cool toy Or ideation source In more of A professional tool.

How has to use THE new Mid Road consistent character functionality

THE label works best with previously generated Mid Road pictures. SO, For example, THE workflow For A user would be be has First of all generate Or to recover THE URL of A previously generated character.

Let's go to start Since scratch And say We are generator A new character with This fast: "A muscle bald man with A pearl And eye patch. »