Adobe turns vocal cues into cinematic sound effects

The Generate Sound Effects feature functions as an intuitive extension of a traditional editing timeline. By recording a "clip-clop" sound alongside a video of a walking horse, users provide the AI with both a rhythmic blueprint and a descriptive text prompt, such as "hooves on concrete." The system then returns four distinct audio variations. While the technology stops short of generating speech, it excels at capturing impact noises—like snapping twigs or closing zippers—and ambient atmospheric textures.

Beyond audio, Adobe is upgrading its Firefly Text-to-Video generator with deeper structural control. Users can now utilize Composition Reference to mirror the framing of existing footage, while keyframe cropping allows for the generation of video transitions between specific start and end frames. These updates include visual presets like anime and claymation. Although early demonstrations of these presets have met with mixed reception, Adobe’s broader strategy involves integrating these fine-tuned controls with third-party models. Generative AI lead Alexandru Costin confirmed that the company intends to extend these capabilities beyond their proprietary software, positioning Adobe as the primary interface for creative workflows as the market for generative models becomes increasingly crowded.