The mid-journey has come a long way. As one of the original AI image generators, popular services are no longer discordant. We expanded our offer with new features and models, including the latest 7th generation image models. Currently, Midjourney is participating in a pool of AI companies supporting the AI video generation.
As a resident AI creative software reviewer at CNET, I knew I needed to try out the V1 video model. I was most interested to see whether the creative and detailed quality that I enjoyed with AI images would be translated into new videos. What I found was the classic Midjourney: a fun result, with a generic AI foible wrapped in an intuitive and quick interface.
Midjourney is a competitive option for AI enthusiasts in an increasingly crowded market. But before you pull out your wallet, you need to know where the cracks and pitfalls are. Here's how my experience with Midjourney has progressed, including the failures and successes of the videos I've produced.
Good, bad, physically opposed nature of Midjourney's videos
Midjourney offers a generation from image to video. You can upload and animate your own images with or without additional text prompts. This is great for constraining the model to the appearance of one image – reducing the chances of hallucination in the wild – but I hope that the company will eventually add a video to video pathway for us who prefer to have ideas with text rather than images.
We first generated AI images to test and stretch Midjourney's artistic abilities. I then asked to animate them with instructions for specific movements or additional elements moving the frame. The result, like any AI creative service, was a hit or miss.
First of all, what I liked: Midjourney videos were easy to make, easily upscale and turned into the usual colorful and whimsical style. This was a quick process, usually about 1-2 minutes in a batch of four variations per prompt. You can adjust some settings, including dimensions, before running the prompt. My favorite tool was the new slider that could add to the weirdness of the video. The video is capped in 5 seconds and has no watermarks or audio.
Bad things now. Midjourney is still struggling with quick compliance or creating exactly what you want. This is something I also encountered with that image generator. For example, in one video, I asked to move the roots of a tree and add a fairy floating above it. I only had new elements, but my tree roots remained static. Text generation is still messy and readable. (To be fair to Midjourney, Openai's new image generator set a fairly high bar.) My dream of revealing the phrase “I'm over It” to the butterfly never came true.
Object persistence is also a struggle. Models may understand and follow the laws of physics, and other overlapping elements of time may disappear and reappear at random. To get an example of Michael Jordan dunking over LeBron James, he needed a lot of generation without his hands passing through the hard hoops or breaking the net. And yes, I was able to generate a semi-recognizable version of the celebrity without using their names at my prompt. The AI moderator kicked occasionally.
It is possible to add an object after the image is created, but in my experience it was rather inaccurate and frustrating. I got the best results when I created the perfect image (or grabbed my favorite from my camera roll) and told Midi Joanie to animate it. I know that adding too many moving parts to a project is a normal disaster, but I expect it to be better if that project is making videos.
How Midjourney stacks up with Veo 3 and its competitors
The best AI video service for you is to depend on your project and the quality level you need to complete it. Amateur creators and midi Joanie enthusiasts will find familiar ground and a lot of fun, but many AI quirks mean that mid-journeys are too short to be useful for professional creators.
One obvious place where Midjourney's videos aren't up to their competitors is to use audio. If you want audio generated and/or synchronized to AI to use AI video clips, you should use VEO 3 or Adobe Firefly. Midjourney's V1 does not have that feature, but it may change in future updates.
Other things to consider when choosing Midjourney and other services are whether you already have the assets you want to animate, or if you want to brainstorm/start from scratch. Midjourney is an excellent AI brainstorming partner, and its creativity helped spark new ideas. This model can easily animate existing images, but the more complicated instructions made it difficult to realize my exact idea. In such cases, patience – and sufficient time of refinement and editing prompts – will become your best friend.
Midjourney pricing, availability and privacy
Midjourney is a paid service available through its website and Discord. That subscription starts at $10 a month. This is cheaper than Google's VEO 3 and Openai's SORA, both starting at $20 a month. (Adobe Direfly also starts at $10 a month.) Upgrading to a more expensive plan will result in more work and faster generation times. This means that you can run multiple prompts at once, and those jobs are faster.
Midjourney's privacy policy states that it can use the information it provides to train machine learning algorithms. The company also states that it can share your information with service providers, third-party vendors and other business partners. All images and videos will be automatically shared in Midjourney's public gallery unless you pony up the stealth mode feature with the Pro Plan at $60 per month or the most expensive megaplan at $120 per month.
Also noteworthy of privacy considerations, Midjourney was recently slapped in a massive copyright infringement lawsuit. Disney and Universal called the AI service “a bottomless hole in plagiarism” in their complaints, claiming that Midjourney allowed users to create surreal images of protected characters like Shrek and Yoda. This case is currently underway and does not affect Midjourney users.
For more information, see How to Write the Best AI Image Prompts and How to Fix the Most Common AI Hallucinations.
