Popular Tools in Artificial Intelligence Image Generation Midjourney recently released an alpha test version of its latest model, Midjourney V7. This update comes almost a year after the release of the last major version, V6, and brings a number of important improvements and new features designed to enhance image quality, user experience, and creative productivity.
Core Update: Image Quality and Text Comprehension Improvements
According to Midjourney officials and its CEO, David Holz, the V7 has been described as "the smartest, most beautiful, and most coherent model yet". The new model shows progress in a number of ways:
- image quality: V7-generated images show significant improvements in texture, detail (especially in long-standing difficulties such as bodies and hands), and overall coherence. Early V7-generated images shared by users demonstrated an impressive level of detail and increased realism. Compared to V6, V7 images are perceived as more natural and less "waxy" or over-edited.
- text comprehension: V7 is able to handle text prompts more intelligently, improving the understanding and execution accuracy of complex or nuanced descriptions.
- Personalization is enabled by default: V7 is the first Midjourney model to have "personalization" enabled by default. Users are required to rate approximately 200 pairs of images to create a personalized visual preference profile, and Midjourney learns from the user's aesthetic preferences and adjusts the generated results to better match their personal style. Users can also choose to turn this feature off.
Innovative features: "draft mode" and voice interaction
One of the biggest highlights of this V7 Alpha release is the introduction of the new Draft Mode. This mode has the following features:
- High efficiency and low cost: Generation is claimed to be up to 10 times faster than the standard model at half the cost.
- Fast Iteration: Ideal for rapid ideation and creative iteration. Users can quickly generate initial ideas, and if they are satisfied with the result, they can click on "Enhance" or "Change" to re-render in full quality.
- voice interactionDraft mode supports voice input. Users can directly speak out their ideas and let the model generate or adjust images in real time, realizing a smooth experience of "creating while thinking", which is called "Dialogue Mode".
- invocation method: Add after the cue word
--draft
parameter to enable it.
Versions and Limitations
Currently, Midjourney V7 Alpha offers two standard rendering modes, Turbo (faster and more costly, twice the cost of V6 regular speed) and Relax (less costly). The standard normal speed mode is still being optimized and will be available in the future.
Note that the Alpha version of V7 does not yet include all features. Features such as image upscaling, editing, and retexturing are still done using the V6 model and are expected to be added to V7 over the next two months.
How to experience and future outlook
To experience Midjourney V7 Alpha, users need to first complete the creation of a personalized profile (by rating at least 200 pairs of images). After that, you can choose to use the V7 version either on the Midjourney website or on the Discord bot (by adding the --v 7
parameter or selected in Setup).
Midjourney promises to roll out new updates and features for V7 models every one to two weeks for the next two months, including new roles and object reference systems.
Despite facing lawsuits over AI training data, Midjourney continues to grow as an independent company with no external funding and plans to expand into new areas such as hardware, video, and 3D object generation. the release of Midjourney V7 heralds another step forward in quality, efficiency, and personalization of AI image generation technology, and is worth keeping an eye on for the industry and users. The release of Midjourney V7 heralds another step forward in AI image generation technology in terms of quality, efficiency and personalization, and its subsequent development deserves the continued attention of the industry and users.