They charge an arm and a leg in gems for video features like Face Swap and Character Swap - and results are garbage. They maintain the facial features of the original video - maybe change hairstyle and clothes, but that's it. However, they'll still charge you the full fee for these useless results. Why not give users a 5-second test run before wasting their money?
SK
SJ K.
The video generator is slow
The video generator is slow, like dial up slow. I got tagged three times for breaking rules I didn't break, including two times for using images of real people when the person in question was an image I made on Mage, glowing eyes and all. What the subscriptions include isn't real clear unless you click on the tiny little dots for the fine print. You can't do anything like edit videos unless you're paying some top tier stuff. Also (though I understand there's going to be some limits on it) but Google is advertising you as a good NSFW image and video generator and honestly, ChatGBT is less picky about their content and they don't actually allow NSFW content at all. I've spent way too much time and money. Subscription has been cancelled and account deleted.
MO
Mageless O.
Unlimited is LIMITED, but quality potential was there, but it is slow. Also has lazy accusatory incompetent support
The service was unusable for 25% of my subscription time due to using the Unlimited too much even without botting or automation.
To address some reviews within two months of my experience, if you use the wrong model such as using Z-Image Turbo for a reference image that you want with a superbly crafted prompt or wanting two leaders with one of them in a wedding dress, the model is going to fail 100% of the time. That is not the model for those prompts. A new user is not going to know that or be impressed with the service if they keep ramming their head up against it.
Where for the angles, Mango 2, GPT-Image 2, or Nano Banana models would have performed dramatically better and given them some hope or achieved it. Albeit the latter two models possibly switching the person a little bit and GPT-Image 2 often distorting the person/character around their time of the experience.
And then due to the censorship on GPT-Image 2 and Nano Banana models, the leaders prompter would've needed to use Mango 2, Mango 3S, or Guava, but probably Mango 2 to get that to work. There will also be occasional nose degradation and torso stretching in Mango 2, which can be problematic.
For the "Unlimited" video generations, the video length is as short as the others mention. It may be just enough to get one motion in, a little quick 3~6 second slogan in, or such.
The video models, WAN leans into the reference image's existing position and ignore most of your prompt and settings. If two people/characters are a little too close to each other, they get touchy and intimate. WAN will be without audio, often in slow motion, and degrade if you try to extend it from an existing video. WAN is the fastest generation of the video models, but do not expect it to really adhere to your prompt or for decent visual quality.
The video models Blueberry 2 has been the best for character consistency and prompt adhering and overall was the least censored so if you want to have a character punch someone in the face, it has the chance to look like a real punch, but takes a long time to generate and only a few seconds in duration. The other models are going to do laughably poor at violence or ignore prompting for "sensual" themes. Or if they are decent at prompting, they can be censored (such as VEO).
Raspberry and Cherry are horrible for prompt adherence and respecting the character (reference image). You'll get that models interpretation of your character and in one of the styles that the models prefer over what style you prompt it for. The saving graces of either model is that they don't result in as many malformed body parts. The characters they do show will be consistent throughout the clip more often. And Cherry is capable of some cinematic shots, great special effects, and lip syncing with audio you provide. But Cherry also often tossed away a video generation without telling me why and would let the same image and prompt go through the next time. So you may end up with only a few videos every hour.
Peach and Kiwi generate videos faster, but quickly lose your character's features if they start moving or turning. The models tend to hallucinate and create malformed body parts or impossible physics a bit more often.
Kiwi is a little better at respecting an action prompt, but horrible with dialogue. Peach is a little better with dialogue, but sometimes speaks Mandarin even with prompts toward speaking English.
There are some other Apps that they advertise such as face swapping, outfit swapping, character generation, or muse generator. The results are little different than if you prompted it yourself. For example, Muse app just scrapes photos of celebrities and Instagram, then encourages you to change a feature of them to make an original character.
UNLIMITED is LIMITED.
Pro Plus and Max have a higher chance of getting you banned for botting. You can queue up 10 generations with those subscription tiers. If you use the service across the day, you will be temp banned for botting as having using the service for 24-hours straight.
This isn't difficult to do when queueing up video generations before sleep, when you wake up, during a lunch break, and after home from work. What is 15 minutes of prompting across the day manifests as 24-hours of system use and a 48-hour ban for botting.
Customer service took 24-48 hours to respond. They insinuate you're botting and lying about it, can't fathom their screw up, don't unban you most times, and suggest using the Unlimited service less.
When banned, you can't easily access Stripe to unsubscribe. Unsub first to avoid complications.
I'd understand and accepted if they banned me for the sensual themes that were against Terms of Service or for IP infringement. But the false advertising of Unlimited and being down 25% of the subscription is very, very disappointing.
HW
HW
Went down hill badly.
Honestly, this used to be my favorite site. Then sometime ago they changes the format of the site and the models just got terrible. I went from being able to create wonderful works of art, to stuff that just looked terrible. I left the site because of the revamp. I have been back on a few times, but it is just not competitive with other art sites now.
I logged on today, in fact, and it is just not intuitive like it used to be and the quality of art is just not as good as the legacy Mage Space was.
Response from mage.spaceJun 22, 2026
Hi HW, I'm sorry to hear that you're not getting the same experience from Mage that you've enjoyed previously. Without diminishing your experience, Mage has been revamped with tons of models and new features, so it's peculiar that your experience now feels lesser than before.
As for the feeling we get when a site evolves and we loved the previous iterations of it... I wish I could be more helpful there and it is true that Mage as grown a lot in 2026. I'm happy to continue the convo here or at [email protected] so we can find and understanding and improve your experience. Thanks, Jason.