OpenAI Launches Images 2.5: Controlled Editing and Two API Models

Key Takeaways

1OpenAI launches ChatGPT Images 2.5, focused on controlled image editing

2Two API models, Flare and Sunburst, are offered; Flare boasts 50% lower latency

3Sketch, Templates, and prompt sharing enhance the ChatGPT interface

Why it matters : Precise editing and fidelity to reference images are becoming central in generative AI tools, marking a shift in usage.

OpenAI introduces ChatGPT Images 2.5 with a clear focus on editing. Two API models, Flare and Sunburst, accompany the update, with a reported latency reduction. On the interface side, Sketch, Templates, and prompt sharing expand the use cases.

Two API Variants and a Reported 50% Latency Reduction

OpenAI offers two API models for Images 2.5. GPT-Image-2.5 Flare brings improvements in quality, editing, and speed, becoming the default option for most applications. GPT-Image-2.5 Sunburst targets high-end visual use cases requiring more advanced editing control, with longer generation times accepted. According to OpenAI, Flare generates higher quality images than GPT-Image-2 and reduces latency by half. Sunburst is presented as suitable for creative workflows and advanced editing.

Sketch, Templates, and Prompt Sharing Arrive in ChatGPT

New tools are integrated into ChatGPT to facilitate work on images. Sketch allows users to draw a rough idea serving as a visual guide, such as a room layout, an outfit, or a sketch, and then specify the desired style. Templates provide starting points for common formats like posters or products, avoiding the need to start from a blank prompt and making it easier to create visuals such as posters or portraits. Prompt sharing associates each image with the prompt that generated it, allowing other users to reuse the same idea with their own images and details.

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Official Demonstrations Emphasize Fidelity and Targeted Editing

Official presentations prioritize the use of reference images. They highlight the ability to maintain subject identification, lighting, textures, and distinctive features, even when modifying context or style. One example relies on a reference animal image to illustrate how to modify an existing subject by applying a different visual treatment. Other demonstrations include a reference-guided photo booth portrait and a composite party scene with multiple people, illustrating the need to preserve multiple subjects and their visual relationships during modifications.

Images 2.5 Defines Itself as an Editing Tool Integrated into ChatGPT

ChatGPT Images 2.5 represents the latest version of OpenAI's image generation model, now present in ChatGPT. This evolution emphasizes controlled editing, promising increased detail, more realistic rendering of light and textures, and improved fidelity to reference images. OpenAI highlights the robustness of successive modifications and indicates a generation latency reduction of up to 50% compared to Images 2.0. The stated priorities include precise editing, fidelity to reference, and consistency across multiple iterations, complemented by new features like Sketch, Templates, and prompt sharing, faster generation, and new API models.