Adobe announced on Tuesday the launch of its latest image generation model, Firefly Image 5. The update also introduces new features for the Firefly website, support for additional third-party models, and the ability to generate speech and soundtracks. A notable addition allows artists to create their own custom image models using their existing artwork.
The new Image 5 model can generate images at native resolutions of up to 4 megapixels, a significant increase from the previous model, which generated images at 1 megapixel and then upscaled them. Adobe also stated the new model is better at rendering human subjects.
Image 5 introduces layered and prompt-based editing. This means the model treats different objects as separate layers, allowing users to edit them using text prompts or tools like resize and rotate. Adobe ensures that these edits do not compromise the image’s details or overall integrity.
The Firefly website already supports third-party models from AI labs such as OpenAI, Google, Runway, Topaz, and Flu. Now, Adobe is expanding this by letting users build custom models based on their personal art style. Currently in a closed beta, this feature allows users to drag and drop assets like images, illustrations, and sketches to train a model on their unique style.
New features have been added to the Firefly website as well. The prompt box can now be used to switch between generating images or videos, select which AI model to use, change aspect ratios, and more. The home page now displays user files and recent generation history, with shortcuts to other Adobe apps that were previously located in a menu.
Adobe has also redesigned its video generation and editing tool to support layers and timeline-based editing. This redesigned tool is currently available in a private beta and will be rolled out to all users in the future.
Two new audio features are coming to Firefly. Users can employ AI prompts to generate entire soundtracks and speech for videos, using models from ElevenLabs. A new prompt creation method is also available, allowing users to build prompts by selecting keywords from a word cloud.
As competitors like Canva integrate more AI into their platforms, Adobe is focusing on new-age creators who regularly use AI in their workflows. The company’s Vice President of Generative AI, Alexandru Costin, explained that the target audience for Firefly is next-generation creative professionals who embrace generative AI. He added that Firefly gives Adobe more freedom to add new features and experiment with the interface, as it is not bound by the established workflows of traditional Creative Cloud tools.