Jimeng AI Releases Version 3.0: Stunning Chinese Text Control Capabilities That Could Put Designers Out of Work 179

Jimeng AI launched a new version 3.0 on April 2, 2025, and is currently in a small-scale internal beta phase. The Chinese text control capabilities of the new version are explosive—this time, it really is putting designers out of a job.

Jimengai version 3.0 has made several improvements in image generation, covering e-commerce product photos, photographic portraits, movie posters, and artistic text images.

What specific improvements does Jimeng AI version 3.0 bring to image generation?

Enhanced Image Refinement

• 2K HD Images: Jimeng AI 3.0 can directly output 2K HD images, delivering exceptional detail performance. For example, in e-commerce product images, details such as the cold light reflection on a smartwatch's metal band and the dynamic heart rate waveform on the watch face are rendered with remarkable clarity.

• Detail Optimization: In photographic portraits, there is a notable improvement in details such as human hands and clothing. For instance, in a 1990s Hong Kong street-style portrait, the fine details of a woman holding an old-fashioned telephone receiver and the movement of her hair in the breeze are rendered vividly.

Optimized Text Generation

• Clear Text Rendering: The text blurring issues seen in previous-generation models have been resolved, and even small text can be displayed clearly. For example, in an artistic text image, the English word "DREAM" is rendered in liquid mercury, with flowing iridescent reflections on the letter surfaces, droplet splashes in some areas, and a deep gray brushed metal background—an Industrial Light & Magic-level material render with a surreal CG quality.

• Text and Image Fusion: The integration of text and imagery is now more natural. For instance, in a movie poster, the title "《Fu Sheng Jie》" is rendered in brush calligraphy, blending seamlessly with the collapsing immortal mountains and golden rune formations in the background.

Precise Composition

• Reduced "Gacha" Attempts: The model follows user prompts more precisely for composition, significantly reducing the number of "gacha" attempts and greatly improving creative efficiency. For example, in a Nordic-style wooden tabletop e-commerce product image, the effect of a serum drop suspended at the bottle's mouth was successfully achieved on the very first generation.

• Complex Scene Composition: The model handles complex scenes with strong compositional accuracy. For instance, in a movie poster featuring a massive alien ship hovering over an abandoned city, the scene showing mechanical tentacles extending from cracks on the ground and the protagonist squad running through the ruins is rendered with clarity.

Enhanced Cinematic Quality

• Cinematic-Level Composition: Generated images carry a cinematic quality that meets the high standards required for movie posters, artistic portraits, and more. For example, in a Chinese-style commercial still life shot on a rosewood tea tray, a wisp of steam rises from the teacup forming the seal script character "tea," with strong light-and-shadow contrast and a Wong Kar-wai movie color palette.

• Film Grain Texture: Film grain has been added to certain images, enhancing their realism and artistic appeal. For example, the 1990s Hong Kong street portrait features Kodak film grain and a Wong Kar-wai-style teal-and-orange color tone.

Overall Effects

• Enhanced Prompt Response: Compared with the previous 2.0 Pro and 2.1 models, prompt response capability has been improved, enabling a better understanding of user descriptions.

• Improved Image Quality: Overall image quality has been elevated, with richer details.

• Enhanced Conceptual Understanding: The ability to interpret concepts has been strengthened, though there is still room for further improvement.

Application Scenarios

• Movie Poster Production: Its cinematic-quality image generation capability enables it to produce high-quality movie posters.

• Design and Creation: It offers designers greater creative space, allowing them to quickly generate images that meet their needs and improve work efficiency.

• Commercial Design: It can rapidly generate images in various styles as design assets or references, boosting work efficiency—examples include e-commerce product photos and advertising posters.

Jimeng AI version 3.0 is currently in the small-scale internal beta phase. Click here now to check whether your account has been granted access to the Jimeng 3.0 grayscale test.