Adobe has released the Firefly Video Model in public beta, bringing the “first commercially safe” AI video generation model to market.

Adobe is pushing the IP safety of the product with Firefly only trained on content that Adobe has permission to use, which includes licensed content from Adobe Stock and public domain content – “never on Adobe customer content.”

The model is available in the Firefly web application with Generate Video (beta) and in Adobe Premiere Pro with Generative Extend (beta).

Generate Video (beta) allows users to generate videos, b-roll, animations and more. Text to Video and Image to Video features allows the creation of 1080p video from text prompts, generation of b-roll to fill gaps in timelines, the addition of atmospheric elements to a shot and development of custom motion design elements. Controls allows for the fine-tuning of a shot with detailed camera settings, like sweeping aerial views, framing from close-ups to expansive long shots, and dynamic movement by selecting from multiple motion paths. Users can also lock in the first and last frames of a shot to preserve visual continuity, keep colours and character details consistent and more.

It supports 1080p resolution to start, with both a lower resolution ideation model for high-speed iteration and a 4K model for pro-level production work coming soon.

Also available is Scene to Image (beta) to create professional quality images from 3D sketches and reference shapes. Using a lightweight 3D sketching tool, creative professionals can transform artwork into high- resolution images, build structure references directly within Text to Image and create precise visual guides using 3D tools. Users can change 3D image angles and perspectives to get the perfect output. With Scene to Image, “creators can seamlessly render production-ready assets from 3D shapes, all with the flexibility and precision required in professional graphic design workflows.”

There’s Translate Audio and Video for the translation of spoken dialogue into multiple languages “while maintaining an authentic voice”. With voice, tone, cadence and acoustic match when translating video content into different languages, creators can also spend less time on dubbing performance and audio mixing. Translate Audio and Video supports translations in over 20 languages.

Adobe is introducing new Firefly Standard and Firefly Pro plan offerings that enable access to the new Firefly web application.

“Firefly is designed for creative professionals looking for unmatched creative control and IP-friendly tools that can be used safely and effectively in production,” said David Wadhwani, president of Adobe’s digital media business.

“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.”

Adobe says that in under two years, Firefly has been used to generate over 18 billion assets globally – an uptick of 50% over the past few months – and evolved from an image generation tool to the most comprehensive generative offering for creative teams. Beyond generating high-quality images, it provides creators unmatched control over their work through professional camera angles and positioning, rich detail and prompt accuracy and the ability to match the structure of style of a reference image in image, video and 3D outputs.