Meta announces 'Make-A-Scene' that can generate high quality images from text explanations and ultra-rough illustrations



In recent years, several AI-based paint tools have appeared, such as NVIDIA's '

NVIDIA Canvas ' that generates landscape images from appropriate graffiti and Open AI's ' DALL-E 2 ' that can generate images from text. Meta, which newly operates Facebook, announced the AI paint tool ' Make-A-Scene ' that can generate more accurate images by combining text explanations and rough illustrations.

New AI Research Tool Turns Ideas Into Art | Meta
https://about.fb.com/news/2022/07/metas-new-ai-research-tool-turns-ideas-into-art/

Greater creative control for AI image generation
https://ai.facebook.com/blog/greater-creative-control-for-ai-image-generation/

Meta's'Make-A-Scene' AI blends human and computer imagination into algorithmic art | Engadget
https://www.engadget.com/metas-make-a-scene-ai-algorithmic-art-130058753.html

Many AI painting tools generate images based on the text you enter, such as 'a cow in the hospital' or 'a zebra riding a bicycle'. However, it is unlikely that an illustration that accurately reflects what you envisioned in your head will be generated just by explaining it in text, and there are complaints such as 'I like a different composition' and 'The zebra is too big compared to a bicycle'. It often comes out.

Therefore, Meta has developed an AI painting tool 'Make-A-Scene' that can generate illustrations that are closer to what you envision by combining textual explanations and simple free-form sketches. You can see what kind of AI painting tool 'Make-A-Scene' is by watching the following tweet embedded movie.



A hand drawing something with a stylus.



On the right side of the frame 'Make-A-Scene with AI', a rough illustration is drawn.



Enter the text 'A painting of an ocropus under the sea, jellyfish in the sky at night'.



Then, from the illustration that I did not understand by itself ...



An illustration of an octopus under the sea and something that seems to be a creature flying in the night sky was generated.



It is also possible to generate multiple illustrations from the same combination of illustrations and text.



Each illustration has a different taste.



In addition to this, Meta publishes multiple illustration generation examples on the official website. Here is an example of a combination of the text 'A colorful sculpture of a cat' and an illustration.



In addition, Meta is sharing access to the Make-A-Scene demo with multiple AI artists as part of the R & D process. This is what AI artist

Scott Eaton actually generated from the text 'Skyscrapers in the desert' and a rough illustration. 'Make-A-Scene offers a level of control not found in other state-of-the-art art-generating AI systems. Textual explanations alone are often constrained and often wander in the dark. Being able to control the composition is a powerful enhancement for artists and designers. '



Meta says, 'Using AI to increase human creativity is a powerful use of technology.' 'It's not enough for AI systems to generate content. AI has the potential to drive creative expression. To achieve this, artists need to be able to shape and control the content generated by the system. '' Make-A-Scene is a generational AI research tool that does just that. '

Related Posts:

in Software,   Science,   Video,   Art, Posted by log1h_ik