Alibaba unveils AI image generator that takes on Midjourney, DALL-E
Chinese e-commerce powerhouse Alibaba is in a race to establish itself as one of the leading names in artificial intelligence (AI). In the last few months, the company has introduced a series of AI tools to compete with its Western rivals. Now, it has launched an AI image generator dubbed 'Tongyi Wanxiang' to take on competitors like DALL-E, Midjourney, and Stable Diffusion.
Why does this story matter?
ChatGPT's imminent rise to popularity was a rude awakening to several Chinese firms, which have been working on AI for a while. Now, they are under pressure to come up with similar or better AI tools. If they don't, China's aspirations to become an AI superpower would be in peril. Especially considering the US's attempts to limit China's access to advanced technology.
Tongyi Wanxiang can generate images and animation
Tongyi Wanxiang was unveiled by Alibaba Cloud during the annual World Artificial Intelligence Conference in Shanghai. According to Alibaba's corporate newsroom Alizila, it can generate images from natural-language prompts. The AI is capable of creating anything from watercolors and oil paintings to animation and 3D cartoons. It is capable of understanding prompts in English and Mandarin.
The AI tool is primarily for enterprises
Alibaba is primarily targeting enterprises with Tongyi Wanxiang. This is a shift from some of the major AI image generators in the world. "With the release of Tongyi Wanxiang, high-quality generative AI imagery will become more accessible," said Zhou Jingren, CTO of Alibaba Cloud Intelligence. The company hopes the tool would be used in various industries, including e-commerce, gaming, design, and advertising.
Tongyi Wanxiang is powered by Composer LLM
In Mandarin, Tongyi Wanxiang means "tens of thousands of images." An appropriate name for an image generator, isn't it? Its prowess is not limited to generating images. It can also apply one image's style to another or create different variations of the same image. Tongyi Wanxiang is powered by Alibaba's large language model called 'Composer,' a text-to-image diffusion model.
The image generator is in beta stage
Tongyi Wanxiang follows Tongyi Qianwen, a GPT-like AI model. It was launched in April this year. Per Alizila, Tongyi Qianwen has received over 300,000 beta requests since its launch. The company would be hoping for the same response to Tongyi Wanxiang. The image generator is also in its beta stage. Alibaba didn't specify whether a non-enterprise version of Tongyi Wanxiang would be available.