Reve

Reve is the image generation startup that earned outsized attention for a focused bet: state-of-the-art image editing through natural language, treating modification — not creation — as the primary problem worth solving. Upload an image, describe the change conversationally ('remove the person on the left, make the lighting warmer, put her in a blazer'), and Reve's models execute while preserving everything you did not mention — identity, composition, style — at a fidelity that independent comparisons consistently place at or near the top of the editing category, ahead of far larger companies' offerings.

The 'half' in its public framing (creating half the image in the world) signals the philosophy: editing is where real creative and commercial work happens — product photography adjustments, portrait retouching at scale, marketing asset iteration — and creation tools historically handled it poorly, either regenerating everything or preserving nothing. Our product-photography tests showed the differentiator concretely: background swaps and object additions that maintained product identity perfectly, the task that repeatedly embarrasses general-purpose editors. The interface is conversationally iterative, with each instruction building on the last.

Honest limits: it is a younger company and platform — API access and enterprise features are maturing; from-scratch generation, while good, is not its center of gravity; and pricing is still finding its shape as the product evolves. For e-commerce teams, agencies, and creators whose daily reality is fixing and adapting images rather than conjuring from nothing, Reve addresses the most commercially valuable, least well-served problem in generative imaging — and currently does it better than anyone.