99% of AI videos are generated by Midjourney's image production.

2024-06-21

Nick St. Pierre's latest post on X platform made a rather bold statement. He claimed that 99% of AI videos created using tools like Luma, Pika, and Runway are based on images originally generated by Midjourney.

"I conducted an experiment using the same prompts on Midjourney, Ideogram, Firefly, DALL-E, and SDXL, and then inputting the images into Runway and Pika Labs. The videos generated by Midjourney and Ideogram had the best results," he said in another intriguing post on X platform.

Another user expressed a similar viewpoint, stating that the best way to use Luma is to generate a key image using your favorite tool, such as Midjourney, and then let the AI animate it. "It still has that AI video feel, with limbs and characters mutating, but you'll be closer to the desired effect," they said.

However, Midjourney is not the only participant in this competition. There are also Stable Diffusion, Adobe Firefly, Canva, and DALL-E.

The Rise of Midjourney

Since its release in 2022, Midjourney has undergone multiple updates, with the latest one introducing the "Character Reference" feature, which generates consistent characters across multiple reference images.

Users have long struggled to achieve consistent images through references, and this feature was designed to address that issue.

Midjourney has attracted users with its ability to generate realistic images. Additionally, the company's free trial policy since its release has also attracted a large number of users.

Midjourney's CEO, David Holz, confirmed that the surge in new users led him to decide to discontinue the free trial service, rather than the viral spread of images. He mentioned "exceptionally high demand and trial abuse" as the reasons in a message on Discord. The "abuse" referred to the viral spread of celebrity and political photos.

However, this free trial resulted in the creation of fake AI images, leading the company to face a collective lawsuit. In the latest news, the company has discontinued its free trial service.

While the absence of the free trial may deter some users, Midjourney's outstanding performance, such as excellent image quality and versatility, makes it a popular choice for users willing to invest in a subscription. Additionally, the generated images are also used for video production.

What about other competitors?

Midjourney has been a strong competitor in the AI image generation market, boasting an impressive 16.4 million active users as of November. On the other hand, Stable Diffusion used to be a formidable opponent but is now facing significant challenges and controversies.

The company behind Stable Diffusion, Stability AI, has found itself in a predicament. Last year, Emad Mostaque resigned as CEO and board member of the company to pursue a decentralized AI initiative. This move came amidst increasing pressure from investors regarding the company's financial situation.

According to Bloomberg, Stability AI has been burning through cash at a rapid rate, with a monthly burn rate of around $8 million as of October 2023. The company attempted to raise more funds with a valuation of $4 billion, but most attempts failed, leading to speculation about the company considering a sale.

While Stable Diffusion claims to have over 10 million daily active users across all channels, as Mostaque mentioned, other competitors like DALL-E also lag behind Midjourney. DALL-E has followers reaching 1.5 million and produces 2 million images per day.

Despite adopting a subscription model from early on, Midjourney's ability to retain customers has been remarkable. This success can be attributed to the platform's user-friendly interface, continuous updates, and high-quality generated images.

The Unparalleled Midjourney

Recently, Midjourney introduced a groundbreaking new feature called "Model Personalization," which allows users to adjust AI algorithms according to their personal preferences, effectively eliminating inherent biases in Midjourney's training data.

Previously, when users provided prompts, Midjourney would generate images based on its own preferences and interpretations shaped by its training data. This could result in output that is not entirely consistent with the user's desired aesthetics or style.

Midjourney's ability to generate highly accurate images brings innovative approaches to image creation. This approach, combined with the updates Midjourney is introducing, is helping other video creation applications improve video quality and produce stunning results.