Turns out Ernie Image Turbo is quite well-versed in anime
Baidu's image model creates detailed anime mashups, sparking viral discussions on AI art capabilities.
Baidu's Ernie Image Turbo, the company's fast image generation model, has captured significant online attention for its unexpected proficiency in anime-style artwork. A viral prompt submitted to the model requested a complex scene featuring characters from Dragon Ball, Doraemon, Sailor Moon, and One Piece in a dynamic, violent interaction. The resulting image accurately rendered Goku's iconic attack pose, Doraemon's pained reaction, and the shocked expressions of background characters Sailor Moon and Monkey D. Luffy, all while adhering to a specified 'studio animation' aesthetic with vibrant colors and high detail.
The viral example highlights a significant, and perhaps unexpected, strength for Baidu's AI in a highly stylized and popular genre. While many AI image models struggle with consistent character representation and specific anime tropes, Ernie Image Turbo's output demonstrated coherent anatomy, recognizable character designs, and appropriate emotional expressions. This success suggests the model's training dataset included substantial anime and manga material, allowing it to parse complex narrative prompts involving multiple copyrighted characters and generate a cohesive, action-packed scene that resonated deeply with anime fan communities online.
- Ernie Image Turbo generated a detailed scene with 4 distinct anime characters from different franchises.
- The model correctly interpreted action descriptors like 'strong punch,' 'launched,' and 'looking shocked.'
- The viral output demonstrates Baidu AI's training on niche, stylized content beyond typical photorealistic datasets.
Why It Matters
Shows AI's growing ability to master specific artistic styles, impacting creative industries and fan content creation.