Adobe Expands Firefly Audio with Universal License

Adobe has made its Firefly audio tools available to the public for generating voices, ambient music, and sound effects. The company emphasizes a universal license and a training policy based on licensed or public data, in a context where platforms label AI-generated content. The options target professional uses with control over expressiveness, starting at $10 per month for a Firefly subscription.

Universal License, Licensed Data Claims, and Content Labeling

Adobe presents audio generated via Firefly as commercially safe, with music automatically accompanied by a universal license. According to a recent survey by the Berklee College of Music, a majority of musicians and video creators have encountered licensing issues, while platforms can penalize the sharing of content containing protected music without permission. The universal license is intended to spare users from fearing personal lawsuits for infringement.

Under pressure from listeners, services like Spotify and Deezer are adding labels to AI-generated tracks, and Suno is doing the same following negative reactions related to training its models on YouTube videos. The difficulty in distinguishing synthetic music from human-produced music by ear reinforces these labeling initiatives. Adobe also states that its AI policy limits training to licensed or public content and data, asserting that it does not use its clients' work to improve its services.

Tools Transitioned from Beta to General Availability

Initially previewed at Adobe Max last year and launched in beta, Adobe's audio tools are now offered in general availability. They extend the trajectory of integrating generative AI within the publisher's flagship programs.

Mark Ethier, executive director of the Berklee Emerging Artistic Technology Lab, notes that AI is already part of creators' daily lives: according to him, one in five respondents uses generative AI at some point in the process, and one-third includes AI-generated audio in their final edits. He also reports a central need expressed by creators: to have expressive control and autonomy, with tools that support rather than remove steps from their workflow.

Expressive Voices, Ambient Music, and Effects: Functional Scope

Voice generation begins with uploading a script, which is transformed into an audio file. The user can choose artificial voices across various genders and ages, translate the output into over 20 languages, and add pronunciation cues for tricky terms. To limit the monotonous effect often associated with speech synthesis, Adobe has implemented emotion labels to guide expressions and claims to have increased training data related to emotions. Jay LeBoeuf specifies that the team initially designed the system to be controlled by adjectives and verbs, and an integrated stack aims for high-quality expressiveness.

On the music side, the tools target background audio: the generated melodies are instrumental and do not aim to produce complete songs. The user composes a prompt by selecting ambiance, genre, and situation, with a free level of detail. They can also let Firefly draft the prompt after uploading a video; the service then generates up to four 30-second sample tracks that may match the content's atmosphere.

For sound effects, generation relies on a prompt or an imported recording. The model can transform a human voice into a deeper roar reminiscent of a dinosaur or monster and align it temporally with the video.

Access via Firefly, Generative Credits, and $10 Monthly Offer

Access to the new tools is through Firefly, Adobe's AI hub. Depending on the Creative Cloud plan and enterprise permissions, this access may be included in the existing subscription.

Audio generations consume generative credits. For dedicated use, Adobe offers a Firefly-only subscription starting at $10 per month.

A Distinct Professional Positioning from Song Generation

While services like Suno can create entire songs in seconds, Adobe highlights targeted tools calibrated for professional use by filmmakers, musicians, and creators. Among the use cases mentioned is the conversion of scripts into narration voices for short formats.

Jay LeBoeuf summarizes this direction by explaining that Adobe is not aiming to produce "wedding music," but rather to resolve concrete friction points in audio creation and editing, with tools deemed useful by their users.