Adobe Unveils Major Generative AI Update: Firefly Image 5 and More
On Tuesday, Adobe announced a significant upgrade to its generative AI tools. The headliner is the new Firefly Image 5 model for image generation, but the company also greatly expanded the functionality of its website, added support for third-party solutions, and introduced new audio tools.
What’s New in Firefly Image 5?
- Higher Quality: The new model generates images with a native resolution of up to 4 megapixels, which is four times greater than the previous version. Previously, images were first generated at 1 megapixel and then upscaled. This promises superior detail and sharpness right from the start.
- Improved Human Rendering: Adobe states that the model is now much better at visualizing people.
- Intelligent Layered Editing: A new feature allows users to edit individual objects within an image using text prompts via a layer system. The model automatically recognizes different elements, letting you modify, resize, and rotate them without compromising the overall integrity and detail of the picture.
Expanded Capabilities of the Firefly Platform
- Personal AI Models: A feature currently in closed beta enables users to create custom image generation models based on their unique style. Users can simply upload their own assets (photos, illustrations), and the AI will learn to replicate their aesthetic.
- Support for Third-Party Models: The platform continues to integrate models from other AI leaders like OpenAI, Google, Runway, and more, giving creators maximum choice.
- Revamped Web Interface: The Firefly website received a more user-friendly interface. Now, a single prompt box allows users to switch between generating images and video, select an AI model, and change aspect ratios. The homepage displays recent files and generation history.
- New Video and Audio Tools:
- The video editor now supports layers and a timeline-based editing workflow (currently in a closed beta).
- New audio features allow users to generate entire soundtracks and speech for videos using AI prompts powered by ElevenLabs’ models.
- The prompt creation process has been simplified with a new method that involves selecting keywords and phrases from a dynamic word cloud.