TechCrunch: Combinator-powered Poly uses AI for art assets TechCrunch

Like generative artificial intelligence chatAnd the DALL-E 2Entrepreneurs who are starting businesses are attracting investors’ attention and looking to make money by developing new business models around them. Poly is one of the most exciting projects that has emerged from space. It allows designers to create virtual assets, including textures, for 3D models, with only text prompts.

Poly is a stock library that looks a lot like Adobe Stock and Shutterstock. However, it is entirely populated by artificial intelligence. Poly is moving forward despite platforms like Getty Images banning AI-generated content for fear of legal backlash.

“Almost everyone knows the all-too-common pain of searching for that perfect symbol, illustration, font, or sound effect online, only to drop and settle for something imperfect. Poly is trying to improve on this,” CEO Abhay Agarwal told TechCrunch in an email interview. A range of creative-focused generative tools will help to improve the company’s performance.

Agarwal, along with Sam Young, co-founded Poly. He was previously a Research Fellow at Microsoft where he published papers on Artificial Intelligence for Societal Impact. Agarwal then started Polytopal, a “human-centric AI” consulting firm that has worked with brands like Spotify, Meta, and Nestlé to develop various intelligent systems. Polytopal, which co-created BeatSaber’s dance floor algorithm, and launched a virtual baking assistant at Toll House that allows users to customize a cookie recipe according to their dietary needs.

“Young and I started Poly in early 2022 out of a shared passion to ‘increase the creative potential in the world’ and joined Y Combinator’s S22 group,” said Agarwal.

Poly’s web-based layout tool is the first to create 3D textures using physical-based rendering maps. “Physically based render” is a type of modeling that displays images in a way similar to the flow of light in real life.

Poly allows designers to describe the texture, such as “tree bark withmoss”, and can optionally provide a reference photo to create a texture for 3D modeling. You can customize the resolution of your model and have normal and inverse maps. These maps are often used in game development to add volume and depth to the surfaces 3D objects.

“Poly trains its generative AI models with many proprietary methods, such as extracting texture information from regular images to increase the learning capabilities of its model,” said Agarwal.

Agarwal was asked how Poly handles sensitive content such as sexually generated imagery or overtly violent imagery. He provided some details, but stated that Poly “carefully, responsibly” inspects its products. He said, “We have not reported any instances of harm so far.”

Poly is looking to compete with traditional asset marketplaces and manual developer build processes. Apart from portals like GameDev Market or OpenGameArt, major game engine vendors such as Unity host and sell assets through their own platforms.

Polly isn’t the only one using artificial intelligence to create game assets. Hotpot and Pixela.ai are direct competitors. They use similar algorithms to create custom wallpapers and animated graphics, as well as other artistic content.

Agarwal claims that Poly’s generative AI outperforms most assets in terms quality of the assets it produces. The jury is still out on this. Poly plans to make its AI service more unique by offering a variety of asset types such as illustrations, sound effects and sprites. It plans to make money by partnering with enterprises, premium integration of design tools and charging a subscription fee to have royalty-free access, including commercial and resale, to assets.

Agarwal claims that Poly’s free service generates unlimited assets for non-commercial purposes. Poly’s professional plan costs hundreds of dollars, but “thousands of” developers use it. The platform has already produced more than 2 million articles.

Investors such as Felicis, Bloomberg Beta and NextView Ventures attracted the momentum, as did AI Grant, AI Grant, Figma Ventures, Figma Ventures, and NextView Ventures attracted $3.9 million in venture capital towards Poly on September’s Y Combinator showday.

“Poly’s clients range from professionals in Fortune 500 companies to freelancers in game design, AR/VR, interior design, architecture and 3D rendering to e-commerce and marketing,” said Agarwal. “Poly has a multi-year runway and can focus on building the best technology possible because the highest quality product is required to stand out and win in this emerging and very active space.”

Credits for the image:Poly

Assuming Poly takes over on a massive scale, it and its AI competitors run the risk of upsetting the artist community — not only because they could threaten livelihoods but because generative AI systems have been shown to be regurgitateThe data they have been trained with (such as existing technical resources). ArtStation, an art community portal, has allowed AI-generated art to be posted on its platform since earlier this year. IndividualsThey protested widely by putting “No AI Art” pictures in their wallets.

Hinted questions regarding the technology remain unanswered. One class action he claimsCopilot, GitHub’s code generator, generates sections of licensed code and does not give credit. This could have implications for AI systems that create art as well as those who use it. Unrelatedly, the US Copyright Office has recently ended copyright protection for a comic created using generative Ai after initially granting it. They stated that only works made by humans are entitled.

Agarwal doesn’t seem to be worried.

Generative AI is often criticized by creators. Many companies in the field are trying to replace creators using automated systems. Poly’s goal has been to empower creators by providing easy access to their assets. design,” Agarwal said. “Building on its current momentum, Poly plans to continue its relentless focus on innovating its generative AI, model training and product development to support more types of design assets and integrate them into designers’ daily workflows.”

Polly currently employs three people, and plans on increasing that number to two or more in the next six- to twelve months.