OpenAI Supplies Image Models to Adobe’s Creative Studio
OpenAI now supplies its image models directly within Adobe’s creative studio.
ChatGPT Images 2.5: Key Features and Improvements
- New API Models: ChatGPT Images 2.5 ships with two API models, one of which is already integrated into Adobe Firefly.
- Image Generation Volume: Over 3 billion images are generated weekly across ChatGPT and its image API.
- Model Enhancements: The new model offers sharper details, more precise editing, and up to 50% lower latency than its predecessor, Images 2.0.
- Developer Options: Two new models are available for developers: GPT-Image-2.5 Flare (default) and Sunburst (slower, offering higher control for production creative work).
Adobe Integrates ChatGPT into Firefly
Adobe has implemented the new GPT-Image-2.5 models within its Firefly AI studio, highlighting a strategic move to position itself as an interface rather than solely focusing on model availability. This integration allows users to leverage leading AI models alongside Adobe’s own tools in one centralized location.
Mutual Distribution and Competitive Landscape
Both OpenAI and Adobe have integrated each other’s services, with over 70 Adobe tools available on ChatGPT and more than 3 billion images generated weekly across both platforms. This mutual distribution creates a competitive landscape where neither company can be certain of capturing the initial point of a creative task, leading to a focus on developing user-friendly interfaces as a key differentiator.
Social Features and Image Sharing
OpenAI has introduced social features to ChatGPT, allowing users to share images along with their prompts. This enables others to run the same prompt on their own photos, fostering a collaborative environment and spreading content organically. For instance, a 1980s headshot prompt has gone viral as a result of this feature.
Additional Interface Enhancements
Other new features include:
- Sketch: Users can now draw within ChatGPT and use the drawing as a reference point for their text prompts.
- Templates: Pre-designed templates for formats like posters and merchandise are available to help users get started quickly.
- Comments: Direct comments can be placed on images, enhancing interaction and collaboration.
Provenance Marking and Metadata
OpenAI emphasizes its commitment to provenance marking with C2PA metadata and invisible watermarking, along with prompt and image checks. They have published a system card for the release, demonstrating their efforts to ensure proper attribution and tracking of generated content at such a large scale.