metamorworks - stock.adobe.com
Google makes video model Veo available on Vertex AI
It's unclear how well the tech giant's generative AI imaging system will compete against Adobe's.
Google Cloud on Tuesday revealed that Veo, its model for text-to-video and image-to-video output, is now available in private preview for customers on Vertex AI.
Vertex AI is Google's generative AI and machine learning platform.
Google first introduced Veo in May as an experimental model. Developed by the tech giant's research unit, Deep Mind, the model generates videos from existing AI-generated images. It also uses text to create short video clips.
Google also revealed that starting next week, Google Cloud customers will have access to Imagen 3, its text-to-image generator, on Vertex AI.
Customers will also be able to access new Imagen 3 features on the Google allowlist. These features include Imagen 3 Editing and Imagen 3 Customization. The allowlist is a register the vendor controls to grant approved users access to some features in Vertex AI.
The updates and changes come as Google continues to push forward following an incident in February when the imaging portion of the multimodal Gemini model produced inaccurate images of historical figures.
Competing against Adobe
While Google has managed to move on from the Gemini malfunction, and Veo is among the most technologically advanced imaging models available, digital media and marketing vendor Adobe still has a bigger share of the generative AI imaging market, said Andy Thurai, an analyst at Constellation Research.
Andy ThuraiAnalyst, Constellation Research
Adobe's image- and video-generating tool, Firefly, is used by many advertisers and creatives.
"Adobe has cornered this market," Thurai said. "It is going to be tough for Google or OpenAI to break into the creative user case that is completely owned by Adobe at this point."
He added that it might be best for Google or ChatGPT creator OpenAI, which has its own imaging system, Dall-E, to find a major partner instead of competing.
One reason customers seem to prefer Adobe over Google and OpenAI is IP safety, said Keith Kirkpatrick, an analyst at Futurum Group. Adobe promises customers that it does not infringe on the copyright or intellectual property rights of human creators of images. It also offers to indemnify users against legal action for inappropriately using original images.
"Both creatives and business line managers who are responsible for content want to make sure that anything that is generated is going to be safe from an IP perspective," Kirkpatrick said. "You certainly wouldn't want to have something where there's a claim because of some piece of content that was generated."
Google provides digital watermarking for every video and image that Veo and Imagen 3 produce.
Google also plans to indemnify Veo outputs on Vertex AI when Veo becomes generally available. The generated output indemnity applies to whenever a Google Cloud generative service or feature goes into general availability, according to the vendor.
Veo and Imagen 3 also have built-in safeguards to protect against the creation of harmful content, Google said.
Even with these safety features, Google's reputation for responsible generative AI imaging is not as strong as Adobe's, Kirkpatrick said. Adobe has assured customers that it trains its models with vetted data -- licensed content such as Adobe Stock images.
"That's a powerful message to the market," Kirkpatrick continued. "Can Google get there eventually? Yeah, I think so."
Veo's advantage
Many creatives in the market will continue to experiment with Veo because it can produce clips of longer than a minute, while Firefly clips are about five seconds long, Kirkpatrick said.
Another advantage to Veo is its ability to understand complex text, Thurai said.
"Most other models have difficulty in understanding and interpreting the text accurately for intention," he said.
One way Google can differentiate itself in the image- and video-generating market is to ally with a major player, possibly from the entertainment industry, said Michael Bennett, AI policy adviser at Northeastern University's Institute for Experiential AI.
"I'm waiting for one of the big players to have a major Hollywood partner -- a major announcement that would be something that I would take as a clear sign, a real tectonic shift in the area," Bennett said.
Esther Ajao is a TechTarget Editorial news writer and podcast host covering artificial intelligence software and systems.