Video appears on six pricing pages as a headline feature and is, in every case, the least finished part of the product.
Six of the fourteen platforms on this site generate video: OurDream AI, SweetDream AI, Candy AI, Secrets AI, Kindroid and SoulGen. Everything else either does not offer it or offers something animated enough to be called video on a feature list without being video in any useful sense.
Four of those six meter generation with tokens, which is the detail that determines whether the feature is usable. Kindroid is the exception among the companion platforms β a flat subscription at $9.99 or $13.99 with no metering β and SoulGen uses hard caps, a fixed number of generations per cycle.
Three to six seconds. No audio. Resolution below what the same platform's still images produce. Motion that is convincing for small movements β a turn of the head, hair, a shift in expression β and falls apart on anything involving hands, a second figure, or a change in camera position.
The failure mode is specific and consistent: the first second is good, the model loses track of the subject's identity somewhere in the middle, and the last second is a different person. This is the same identity-persistence problem that makes still-image consistency hard, except a video shows you the drift happening in real time instead of letting you discard the bad generations.
Text and hands, already the weakest part of still generation, are worse in motion because errors now flicker frame to frame.
On the four token-metered platforms, a video generation costs several times an image. The exact multiple varies and moves, but the shape is consistent: a monthly allowance that comfortably covers a hundred images covers a small handful of clips.
Which produces the specific failure this post exists to warn about. Video generation fails more often than image generation, the failures are only visible after generating, and on some platforms a discarded clip has still spent its tokens. It is entirely possible to spend an evening and a month's allowance and keep nothing.
If you are going to try it, try it at the start of a billing cycle, not the end, and generate two or three clips before deciding whether to spend the rest on it.
Short, simple, single-subject motion with a static camera. A portrait that breathes and blinks. Something that would have been a still, with a small amount of life in it.
Used that way it works, and on Kindroid β where it is not metered β it is a pleasant part of the product rather than a resource to ration. That flat-rate structure is most of why it is the only platform here where we would call video worth having rather than worth trying.
What it is not is a way to produce anything with a narrative, a cut, or more than one person in frame. The category is not close to that yet.
Because it is the newest thing, and in a category where six platforms have converged on roughly the same feature list at roughly the same price, the newest thing is what differentiates a plan.
This is worth naming plainly: video is currently a marketing feature more than a product feature. That will change β it is improving faster than any other part of the stack, and the gap between a 2025 clip and a 2026 clip is larger than the gap between two years of image generation. But it is not there now, and paying more for a plan because it includes video is paying for something you will use twice.
Do not choose a platform for its video. Choose it for chat or for images, and treat video as something you will experiment with for one evening.
If video is genuinely the point for you, the honest answer is that the dedicated video generators outside this category are further along than anything bundled into a companion platform, in exactly the same way that dedicated image generators beat bundled image generation. The pattern in this category is consistent: bundled is always behind dedicated, and it is always more expensive per output.
Advertised entry prices cluster between roughly $6 and $16 a month, but the sticker price is frequently not the real one. Almost every headline figure assumes annual billing; paying month to month typically costs 40β100% more. On top of that, several platforms charge a subscription for access and then bill image and video generation separately through tokens, so heavy use can double or triple the monthly cost.
A token or credit system means your subscription buys access, not usage. Each image, video or voice message draws down a monthly allowance, and once it's gone you buy more. This is where the real cost lives for anyone who generates regularly. OurDream, Candy AI, Secrets AI and SweetDream AI all work this way. The alternative model is a hard cap β DreamGF and SoulGen give you a fixed number of generations per cycle and charge nothing beyond it, which costs more on paper but is predictable.
Every platform in this category calls itself uncensored and every one of them refuses things. Here is where refusals actually come from, which limits are permanent, and what the word is worth when you see it on a pricing page.
Read postThe same generator is almost never good at both styles, and picking by style narrows fourteen platforms to two or three faster than any score does. What each end of the category is actually good at.
Read postSix platforms do voice and only some of them do real-time calling. Latency is the entire experience, voice messages and voice calls are different products, and on token platforms speech is quietly expensive.
Read post
