CogView - DALL-E in simplified Chinese
The multimodal space is heating up, earlier this week, a research group announced their own version of DALL-E which is 4 billion parameters and takes in simplified Chinese and generates images.
If you don’t know what DALL-E is, DALL-E is a model OpenAI published about earlier this year which can generate entire images just with a text description.
You can play around with this Chinese Version of DALL-E at the demo below:
In my opinion, the results are not as good as the original DALL-E paper suggested, but it’s still a fun little open-source model to play around with. Try using DeepL to translate (like my friend Vova suggested to me) and then feeding the translation into the demo to get back your results.