After Google, Meta has taken its shot at an image-generating AI. Meta announced its own iteration of an image-generating AI competing with the likes of DALL E 2 and Google Imagen.
Text-to-image models like these have existed for a while but have recently gotten more popular. Meta’s Make-A-Scene is the latest AI marvel from the company that can create an image according to a user’s prompt.
Make-A-Scene AI is different: says Meta.
Firstly, The popular algorithmic technology, text-to-image production, is on the rise. OpenAI’s DALL-E mini and Google’s Imagen AIs are producing some mind-bending, odd-looking art. Now, Meta has decided to follow suit. On Tuesday, Meta unveiled an AI image-generating engine in hopes of aiding the creation of high-quality digital art.
According to the company, their version of the AI differs from the ones currently available. Text descriptions were frequently used as input by older image-generating AI systems, and the outcome was unpredictable.
However, this is not the case anymore, thanks to Make-A-Scene. Although, It exemplifies how people can use various elements, including text and straightforward illustrations, to communicate their visions more precisely.
Secondly, according to a report from Engadget, a lot of effort goes into producing an image, based solely on a prompt. You can opt for any sentence, and the complexity of it will decide the result.
Modern state-of-the-art AIs can produce photorealistic images of about any statement you feed them; It’s the result of efforts in training ML models on increasingly expansive, high-definition image sets with carefully vetted text descriptions. However, till now, users had no control over the specific details of the produced image.
Are text-to-image models perfect?
Lastly, Make-A-Scene achieves its goal by combining user-created sketches with its text-based image generation and successfully produces a 2,048 x 2,048p image. Although Make-A-Scene might appear as a tool for artists, Meta claims it’s built for everyone.
“Make-A-Scene isn’t just for artists — we believe it could help everyone better express themselves,” the post said. Even though these systems can produce almost anything with a statement, they are still not perfect. Did you know that the DALL E mini is obsessed with women in saris?
What do you think about Meta’s take on an image-generating AI? Comment down below.