Midjourney launches V7 with improvements in image quality and new features

Midjourney has made an exciting announcement with the launch of its new image model, V7, which is beginning to roll out in alpha mode. This update brings with it a significantly superior image quality, with more beautiful textures and improved consistency in elements like hands and objects. But that’s not all, as the model includes two modes: Turbo and Relax, where the former has a higher cost, more than double that of the standard V6 mode.

One of the most striking features is the new eraser mode, which allows you to generate images at half the cost of the standard mode, but ten times faster. You can access this mode in two ways: by manually activating it or using the flag –draft alongside your prompt.

Customization and Performance Improvements

The images generated in eraser mode are of lower quality compared to the standard mode. However, if you like something you generated, you can click on enhance or vary to get a high-quality version. This model is also compatible with tasks like inspiration boards and style reference codes (SREF), while for more advanced tasks like scaling and retexturing, the V6 models will continue to be used.

An interesting aspect is that to use V7, you must unlock your customization, which is enabled by default. This means that the AI will learn your style and preferences over time, generating images that better match what you like. To achieve this, you need to classify at least 200 pairs of images, a process that takes about five minutes. Once completed, you can enable or disable customization as you wish.

Since its creation in 2022 by David Holz, Midjourney has evolved considerably. In its early days, the platform operated entirely through Discord, but now you can access its official website, although you will require a subscription if you are not a Discord user.

Midjourney is not alone in this race; it faces fierce competition from AI labs like OpenAI, which recently launched a highly praised image generation tool, capable of rendering text, something that has been a weak point for many image generation models. This competition promises to make the future of AI image generation even more exciting.