In the award-winning animation “The Wild Robot,” our robot Roz realizes she must activate learning mode in order to understand and communicate with the world around her. Her system not only had to learn visual elements, but also incorporate and interpret sounds as language. 

Recently, Adobe’s generative AI platform, Firefly, has incorporated sound generation through prompting, and that expands today in a new announcement.

Generate Sound Effects (beta) was available previously in the Firefly system. Users can use their own voice to control the style and pacing of the performance, so to speak (pun intended). The sounds are generated quickly, a fraction of the time spent creating visual assets or using generative AI. 

This morning, I prompted Firefly to generate a robot welcoming you to this site… and the resulting audio was creepy enough to prompt nightmares (pun intended). So, word recognition does not appear to be in Generate Sound Effects through that route. The Text-to-Avatar system was more appropriate for that prompt anyway. Perhaps we’re not yet as sophisticated as Wild Robot Roz, who transitions in Chapter 20 of the book: “…she no longer heard animal noises. Now she heard animal words.” 

As of this morning (7/17/25), for me, “Text to Sounds (beta)” and “Voice to Sound Effects (beta)” are now listed as options for sound generation in the Firefly web application. Generate Sound Effects used to be listed as an option on the main landing page, but the revamped Generate Sound Effects is where you actually land after you click through either Text to Sounds or Voice to Sounds. 

Moooooving along….the system also promises smoother movement and transitions in the Firefly video model.

Also launching this morning is the ability to generate images and videos through several integrations of other AI video models through Firefly. Of particular note to video editors and photographers, Topaz Labs’ popular image and video upscalers are headed to Firefly Boards (Adobe’s Firefly-powered mood board app that launched in June). Do we have a future with Topaz Upscalers further integrated into Firefly or…dream of dreams… directly in Premiere?? “Cannot compute!”

Google Veo 3 with audio is a new integration in Text to Video in the Firefly App and runway’s Gen-4 video is a new integration into Firefly Boards. Also coming soon to the Boards is Moonvalley’s Morey, and Pika 2.2 is headed to the app.

No matter which model is used, Adobe has made clear that the content made through Firefly will not be used to train AI models. All of the integrations can be accessed directly through Firefly. While Content Credentials are applied to content created in Adobe Firefly, which appears to still be the only commercially safe AI Video generator on the list (and when you select your generation option, Adobe has made this abundantly clear).

Non-Adobe Models can also be turned off by an organization, as seen here in my higher educational setup.  

Also available in Firefly video this A.M. is composition reference, keyframe cropping, and style presets within the Firefly app. Text to Avatar (beta) comes to the App as well. 

More information is available through Adobe’s announcement.