Adobe introduced video generation features for its AI platform, Firefly, just before its Adobe MAX event. Starting today, users can explore the video generator on Adobe’s website or try the new AI-powered video feature, Generative Extend, available in the Premiere Pro beta.
On the Firefly site, users have access to two models: one that converts text to video and another that converts images to video, both capable of generating up to five seconds of AI-created footage. (This beta version is free but may have speed limitations).
Adobe claims that Firefly is trained to create animated and photorealistic content according to the specifications provided in prompts. Moreover, Firefly can generate videos with text, something that has historically been challenging for AI image generators. The interface includes options to adjust camera movement, its intensity, angles, and shot sizes.
In the Premiere Pro beta, users can experiment with Firefly’s Generative Extend feature, which adds up to two additional seconds to a video clip, maintaining the camera and subject’s motion. The background audio also extends, marking the first public demonstration of Adobe’s secretive AI audio model. However, the tool will not recreate voices or music to avoid legal issues with record labels.
During pre-launch demonstrations, Firefly’s Generative Extend feature proved more effective than the text or image-to-video models, showing practical utility. While Firefly’s results aren’t as polished as competitors like Runway’s Gen-3 Alpha or OpenAI’s Sora, Adobe has prioritized AI editing tools to meet its user base’s needs.
Adobe faces the challenge of balancing technological innovation with the expectations of its creative audience. Many creatives fear that AI tools might replace work traditionally done with physical tools, so Adobe has focused its AI on enhancing and complementing existing workflows rather than replacing them.
Adobe’s VP of Generative AI, Alexandru Costin, emphasized that users prefer AI to help them edit and expand their resources rather than create new content from scratch. This reflects Adobe’s successful approach with Photoshop’s Generative Fill feature, which has been widely used due to its ability to improve workflows.
Aware of the creative sector’s concerns, Adobe claims to be collaborating with artists and photographers, compensating them for the content used to train its AI models. Despite this, uncertainty remains about the long-term impact of these tools.
Costin argues that generative AI will increase demand for personalized content, thereby boosting the need for creative work. He suggests that those who embrace these tools will be able to produce much more without sacrificing their time.
Firefly also automatically embeds watermarks in AI-generated videos, helping identify the content as AI-created, though these marks won’t be directly visible to viewers. Additionally, Adobe states that Firefly is designed to avoid generating problematic content, such as violence, drugs, or political figures, ensuring a “safe for commercial use” environment. With public access to this technology, time will tell if it lives up to these expectations.
By: Nestor Castillo, ForAllTechNews Director