IT之家 3 月 27 日消息，科技媒體 WinBuzzer 昨日（3 月 26 日）發佈博文，報導稱新銳 AI 圖像生成工具 Reve Image 1.0 以低價高質為賣點，Officially to MidJourney and Flux 發挑挑挑戰。

The model features accurate understanding of user instructions, superior aesthetics, and text rendering capabilities, and is currently available for free trial, with APIs likely to be available in the future. Third-party tests have shown that the image generation quality has surpassed that of competitors such as Midjourney v1.0.

The tool supports natural language modification of existing images (such as adjusting color and perspective), and can upload reference images to achieve style transfer, in addition to breaking through the pain point of text rendering in the industry, and benchmarking its performance against Ideogram, which focuses on LOGO design.

In the third-party platform Artificial Analysis's "Image Generation Quality" rankings, Reve Image beat competitors like Google Imagen 3 to take the first place, with reviews particularly recognizing the clarity of the text within the images it produces. IT Home attaches the relevant screenshots as follows:

Co-founder Michaël Gharbi emphasizes that the team aims to build a "semantic expression layer that is common to both humans and machines", rather than just looking for visual realism. Engineer Hunter Loftis et al. point out that the current model, like earlier LLMs, suffers from a lack of logical consistency.

In terms of cost, the basic package has 500 free generation credits, which will be given away upon registration, with an additional 0 free chances per day, and long-term users can enjoy an ultra-low unit price of $0/0 tickets.