🎨 Reve Releases 4K Model Reve 2.1
Reve has introduced the updated Reve 2.1 image generation model. The new version features improved understanding of complex prompts and high-quality rendering of non-English text. The architecture has become staged: first, composition planning is performed (positions of objects and inscriptions), followed by the final rendering.
🌍 The transition to a staged architecture increases compositional accuracy and generation control. The model's success in the Text-to-Image Arena (2nd place with 1306 points) confirms Reve's competitiveness.
👤 Users receive a tool with more precise visual storytelling and the ability to correctly display text in various languages.
Source 1: https://x.com/reve/status/2075248950756716747 Source 2: https://x.com/arena/status/2075251593277300787