Adobe is unveiling a slew of new artificial intelligence (AI) tools that can edit images or video and generate background music using text-based prompts (promt·instructions).
Ahead of its annual creative conference "Adobe Max (MAX)," to be held Oct. 28–30 (local time) in Los Angeles, the United States, Adobe unveiled new AI-based features during an online press briefing on the 24th. Over the past few years, Adobe has advanced image, video, and audio generation and editing capabilities using AI technology. This year, it added features that let users type what they want as if chatting with AI chatbots like ChatGPT and easily create or recompose works in Firefly, Adobe's Generative AI tool. The company integrated this AI agent–based conversational interface across its major services, including Adobe Express, Firefly, and Photoshop.
Alexandru Costin, Adobe's executive vice president in charge of Generative AI, said, "Creators don't just want to generate new content; they want to quickly and easily tweak content they shot or made themselves," and added, "Accordingly, the newly announced 'Firefly Image 5' model introduces a 'prompt-to-edit' feature that lets you edit images or video with simple text instructions."
In a demo video shown immediately afterward, Costin demonstrated editing a photo of noodles taken at a popular restaurant. When typing in the chat window, "Change the table mat color to blue," the yellow mat turned blue, and when instructed, "Change the car key next to the noodles to a chili oil side dish," the car key disappeared and a small dish containing red chili oil appeared.
Images generated by prompts are provided at 4 megapixels, and content modified through prompt-based editing is provided at 2 megapixels. Adobe said they are produced in high quality and high resolution so they can be used immediately for practical work such as advertising and film production.
A "layer-based image editing" feature that enables more precise image editing is also being added to the new Firefly model. Based on generative compositing technology, this editing feature recognizes image sources as layers and separates them so they can be moved freely. When this feature is applied to the noodle photo shown in the demo, it recognizes the chopsticks placed on the noodle bowl as a separate layer and object, making it possible to detach them and smoothly move them onto the tablecloth.
An Adobe official said, "The newly launched AI-based tools help professional creators handle details down to the pixel level more efficiently and precisely."
A new "Generate Soundtrack" feature that creates background music for video is also being introduced. Entering prompts such as "A peaceful reflective song with ambient electronic style for a travel documentary" generates an instrument-based soundtrack that can be used for commercial purposes such as films, drama, and advertisements. Costin said, "To make soundtrack generation easier and more intuitive, when users upload a video that needs background music, Firefly automatically recommends the mood, tempo, and energy level of music that would suit the video."
In audio, Adobe is also launching a "Generate Speech" feature in collaboration with voice AI company ElevenLabs that can create voiceovers with a variety of voices, timbres, and intonations.
Adobe is also expanding AI features across its Creative Cloud. It improved the performance of Photoshop's "Generative Fill" and added Google's AI image service "Nano Banana" and Black Forest Labs' "BFL" model for more precise retouching. In the Lightroom program, it is introducing an AI-based culling feature that automatically selects the best images from photos taken by photographers.
Costin said, "The goal is to drive innovation in the creative process through AI," adding, "We will help creators worldwide easily make 'scroll-stopping content.'"