Every app that leads with visuals is either shipping video or talking about shipping video. Having spent real money on the ones that have it, my honest read is that the feature is genuinely new capability and genuinely not ready to be the reason you pick an app.
Both of those things are true at once, and the marketing only tells you the first one.
What works
The output can be startling. Watching a scene you described actually move is a different experience from seeing a still, and the gap is larger than the gap between a good still and a mediocre one. When a generation lands, it lands.
That is a real technical achievement and worth saying clearly, because the rest of this piece is more critical.
What does not
It is slow. Slow enough that you stop waiting and go do something else. That single fact changes what the feature is. In-chat image generation works because the picture arrives inside the conversation, while the scene is still live. Video arrives after the moment has passed, which makes it a separate activity that happens to share a login with a chat app.
The hit rate is inconsistent. You will regenerate. Sometimes several times. And because every app offering video runs a credit system, you pay for the failures at the same rate as the successes. The effective cost per usable clip is therefore some multiple of the advertised cost per generation, and no pricing page tells you the multiplier.
It pulls resources from everything else. This is the pattern I find most telling. In my benchmark of 22 apps, the video-first apps are not strong on conversation or memory. Xotic AI scored 7.0 overall, the lowest in my sample, with a 6.0 on memory, while doing video that almost nothing else does. That is not a coincidence. Building video well is expensive enough that something else gets less attention.
The exception
OurDream is the counterexample and worth naming, because it scored 9.5 while also doing video. It is the app I point at when someone asks whether video and a competent companion can coexist, and currently it is a fairly short list with one name on it.
What that suggests is that the problem is not video itself but funding. An app with enough resources can do both. Most cannot, and the ones that cannot are shipping video anyway because the category is in a feature race.
Should you wait
If video is the specific thing you want, there are two apps worth your time and one of them is clearly better. Go and look, with your expectations set to “impressive but slow” rather than “the app moves now”.
If video is a nice-to-have and conversation is the real reason you are here, wait. Buying a weaker companion for a feature you will use twice a month is the most common expensive mistake in this category, and the feature will be better and cheaper in a year.
The bigger picture
The thing I keep noticing is that these races are always won by whoever can afford to lose money on compute longest, not by whoever has the best idea. Video is expensive per generation in a way text simply is not, so the apps that survive this phase will be the well-funded ones rather than the clever ones.
That is worth knowing as a buyer. When an app you like ships video, the question is not whether the feature is good. It is whether the company can afford to keep running it, because a feature that gets quietly restricted after you subscribed is worse than a feature that never shipped. Pricing across the whole tested field is in my pricing report.