Ernie shows some strength in infographic (but yes, in photorealism I still prefer ZIT)
Baidu's Ernie model generates strong infographics but still trails Midjourney in photorealistic image quality.
A viral comparison on social media platforms, originating from Reddit user Zealousideal_Dog8817, has put Baidu's Ernie AI model back in the spotlight. The analysis used a series of creative prompts borrowed from the popular 'nano-banana' image generation trend to test different AI capabilities. The results show that Ernie, China's flagship AI model developed by Baidu, demonstrates surprising strength and coherence in generating infographics—complex images that combine data visualization, text, and icons to explain concepts.
However, the same comparison reveals a continued gap in a key area: photorealism. When tasked with creating lifelike images, users noted a clear preference for models like Midjourney (referenced in the original post as 'ZIT'). This indicates that while Ernie has made significant strides in structured, logical visual generation, it still trails leading Western models in the nuanced, texture-rich domain of photorealistic art. The test underscores the specialized progress of different AI systems in the highly competitive generative image landscape.
- Baidu's Ernie AI shows competitive performance in generating structured infographics from complex prompts.
- User analysis reveals Ernie still lags behind models like Midjourney in achieving high-quality photorealism.
- The test used the viral 'nano-banana' prompt set, a community benchmark for creative AI image generation.
Why It Matters
Highlights the specialized strengths and ongoing competitive gaps between major AI models in the fast-evolving generative art space.