Adobe’s text- and image-to-video AI generator has been released for anyone to try online. Generate Video is available starting today in public beta following a limited early access testing period last year. The beta tool can be accessed via the re-designed Firefly web app, alongside new image generation and translation capabilities, and AI credit subscription tiers for creators.
Adobe started launching tools powered by its generative AI Firefly Video Model in October, starting with the Generative Extend tool in beta for Premiere Pro that can be used to extend the end or beginning of footage. The Generate Video tool rolling out today — over two months after OpenAI launched its own text-to-video generator, Sora — adds some minor improvements since it was first teased in September.
Generate Video consists of two features: Text-to-Video and Image-to-Video. As those names imply, Text-to-Video allows users to generate footage using text description, while Image-to-Video lets you add a reference image alongside the prompt to provide a starting point for the video. Generate Video includes various options that refine or guide the results, such as simulating styles, camera angles, motion, and shooting distances.
Video is now output in 1080p at 24 frames per second, up from the original 720p quality. Both Text-to-Video and Image-to-Video take 90 seconds or longer to generate clips at a maximum length of five seconds — shorter than the 20-second duration available to Sora users. Adobe says it’s also working on both a faster, lower-resolution “ideation model” and a 4K model, which are “coming soon.”
Adobe has also updated the Firefly web app that hosts many of its generative AI tools. Alongside sporting a new UI, it now integrates with Creative Cloud apps including Photoshop, Premiere Pro, and Express, making it easier to move and edit AI-generated assets. And because Firefly is trained on public domain and licensed content, it’s safe for commercial use. Adobe even describes its Generate Video tool as “production-ready” to entice users who want to use AI-generated videos in films without the risk of violating copyright protections.
Adobe faces growing competition in the AI video market. In addition to Sora, Google is testing the second generation of its Veo AI video model, which looks more impressive than OpenAI’s model judging by early demo examples. ByteDance and Pika Labs have also recently announced new video-focused generative AI tools. Adobe’s main advantage is Firefly’s commercial viability, but it will still need to keep up with the quality and features its competitors are offering.
Two additional tools will also be available in public beta on the Firefly web app starting today — but these aren’t free to use. Scene to Image lets users create their own references for AI-generated images using built-in 3D and sketching features — seemingly built on the “Project Scenic” experiment that Adobe announced in October. The Translate Audio and Video tool is pretty self-explanatory, allowing users to translate and dub audio into over 20 languages while preserving the original speaker’s voice.
Adobe is launching two new Firefly subscription plans which provide credits that can be spent to use Adobe’s Firefly models. Firefly Standard starts at $9.99 per month for 2,000 video/audio credits and provides up to 20 five-second 1080p video generations. The pricier Firefly Pro plan starts at $29.99 for 7,000 credits and up to 70 five-second 1080p video generations. A notable perk is that both plans include unlimited access to Firefly imaging and vector features.
