Midjourney, a leader in the AI image generation sector, announced plans to launch a new web-based AI image processing tool next week. CEO David Holtz revealed that the tool will officially debut in early next week.
This tool will feature two main capabilities: AI-driven editing of uploaded web images and the ability to retexture objects. Users can effortlessly alter the color and details of objects through simple text descriptions while maintaining their original shapes.
However, due to potential misuse risks, Midjourney has decided to initially offer access to a limited group of existing users. The company is conducting a comprehensive survey within its community to determine which users qualify for early access.
On the Discord platform, Midjourney candidly stated, "We are unsure how to precisely limit the deployment of this feature, so we aim to gather extensive feedback through surveys." To prevent misuse, the company plans to implement a dual review system involving both manual and AI-based checks.
Previously, Midjourney faced significant challenges with image misuse. In March 2023, the company had to temporarily suspend free services due to the rampant spread of fake images on social media. Free services were reinstated in August of this year.
Video Models, 3D Technology, and Hardware Development
In addition to the aforementioned image processing tool, Midjourney is actively developing video generation models. In an interview with the company’s professional Nick St. Pierre, CEO David Holtz shared unique insights into the video model development process.
The company plans to offer users multiple low-resolution options for selection and enhancement, rather than a single high-resolution video. This strategy aims to provide users with greater flexibility and control. Holtz specifically noted that slow rendering speeds could pose a significant obstacle in the creative process.
Currently, the company is striving to find the optimal balance between speed, cost, and quality. Factors such as reducing resolution or frame rate may introduce new technical challenges. To address this, Midjourney plans to train multiple versions of the model to identify the best possible solutions.
