The Adobe Firefly Video Model is now being released in public beta, as it was perhaps envisioned. Adobe insists they are bringing the “industry’s first commercially safe AI video generation model” to market, in what is essentially the third chapter of its evolution. Last September, Adobe announced this generative artificial intelligence (AI) evolution, which was followed through in October at the Adobe MAX keynote when this was integrated within Adobe Premiere Pro with Generative Extend, also as a beta. The latest unlock, albeit some it residing behind a new subscription, makes Firefly Video Model available for users on the Firefly web application.
Adobe’s urgency in making generative video available to a wider user base, one that goes beyond its creative apps, has its roots in competition that’s making urgent moves too. Irrespective of generative video’s limitations in some cases, Meta announced the VideoJAM AI framework earlier this month, with the idea being to increase coherence in AI generated videos. Late last year, Google detailed the Veo 2 video generation models while OpenAI’s Sora had made its biggest pitch yet by being part of the OpenAI Super Bowl ad. TikTok’s parent company ByteDance’s new Jimeng AI. It may not be outlandish to expect Chinese AI company DeepSeek, which has given the Silicon Valley a few sleepless nights with its frugal approach to AI models, to also have a generative video tool up its sleeve.
The Generate Video (beta) feature is underlined by the Firefly family of creative generative AI models, much in the same way as the Firefly Video Model integration within Adobe’s creative apps including Premiere Pro. The company says this gives creators the tools to generate video clips from a text prompt or image, and the arsenal of capabilities includes ability to choose camera angles for the generation, size of the shot, resolution and aspect ratio, as well as an ability to create professional quality images from 3D sketches. Generations can be further customised with tweaks available for atmospheric elements, special effects such as smoke or fire, as well as custom motion design elements.
Also Read: We aren’t building AI models for the sake of it: Adobe’s Deepa Subramaniam
“Firefly is designed for creative professionals looking for unmatched creative control and IP-friendly tools that can be used safely and effectively in production. We’ve been thrilled to hear from beta customers who’ve found it a game-changer for ideating concepts and producing stunning videos, and we can’t wait to see how the creative community uses it to bring their stories to the world,” says David Wadhwani, president of Adobe’s digital media business.
Firefly Video Model supports 1080p resolution on the Firefly web application for now, but a lower resolution ideation model for high-speed iteration as well as a 4K model for higher resolution generations are “coming soon”. On the Firefly website, you’ll see the new features listed as Text to video (beta), Scene to image (beta) and Image to video (beta), alongside Translate video and Translate audio.
There is a new subscription that users will have to navigate. HT has confirmation that there are three new plans — Firefly Standard ( ₹797.68 per month or ₹7,986.24 per year), Firefly Pro ( ₹2394.22 per month or ₹23,944.56 annually) and a yet to be priced top-tier Firefly Premium plan. The Firefly Standard plan includes 2,000 video or audio credits per month, which is up to 20 of 5-second 1080p video generations. The Firefly Pro plan includes 7,000 video or audio credits per month, which is up to 70 generations of 5-second 1080p video generations.
Also Read: Our AI innovation undergoes careful evaluation and diligence: Adobe’s Grace Yee
Adobe says the Firefly Premium plan is “designed for professionals looking to generate video content on a regular basis will be coming soon to provide even more audio and video capacity for high-volume creators and teams.”
In case you are wondering what happens to existing Creative Cloud subscriptions, Adobe confirms to HT that those will include “limited taste of video and audio generative features” on the Firefly web application. They do not, for now, detail any specifics about the credits available for this tier. That’s one side of the coin. On the other side, no limitations on the Firefly integration with apps such as Premiere Pro and the generative video functionality that’s been available since late last year. For instance, Generative Extend in the Premiere Pro video editing suite, uses Firefly Video model to create footage that can fill gaps in a video b-roll that’s being edited, smoothen transitions or even allow editors to hold a frame for longer to help with more precise editing. That availability isn’t expected to change anytime soon.
“Video is hard. We’ve been working on it for a while, and we weren’t happy with the quality. But our research department has done wonders to handle the 100 times more pixels, 100 times more data, using thousands and thousands of GPUs (or graphics processing units, a computing hardware) to make sure we master this research process,” Alexandru Costin, Vice President of Generative AI at Adobe, had explained to HT during Adobe MAX last year.
Also Read: As the world grapples with deepfakes, AI companies agree to a set of principles
Adobe insists that the Firefly family of creative generative AI models, which has been used to generate over 18 billion assets till now, does not waver from the company’s AI Ethics principles. The training is also only done using content that Adobe has permission to use including what’s in the public domain as well as Adobe Stock media. “Never on Adobe customer content,” the reiterate.
Adobe, through the past year, took the lead in pushing for labels and watermarking of AI generated content, called content credentials, to separate generations from real photos and videos, and has also developed a content authenticity web app that gives creators and users quick controls over labelling their original photos, videos and audio content with provenance data that cannot be altered later.
