For years the visual ceiling in AI companion apps was still images. Good ones, increasingly, but static. That ceiling is starting to lift. More apps are rolling out video generation, and after spending real time with it, I want to give an honest read on where the feature actually stands rather than repeat the launch hype.
The short version
Video in companion apps is real, it is improving, and it is not finished. The best of it is genuinely impressive for something this new. The rest is short, occasionally inconsistent, and expensive to generate. If you go in expecting a seamless moving companion, you will be disappointed. If you go in expecting a promising early feature with rough edges, you will be pleasantly surprised more often than not.
What it looks like in practice
The apps leading here are the visual specialists, the ones where generation is the whole point rather than a side feature bolted onto a chat product. My OurDream review covers the app that has impressed me most on this front, because video is a core differentiator there rather than an afterthought. Being able to get a short moving clip that fits the scene, rather than only a still, changes the feel of the experience in a way images alone do not.
Where it falls short
Two honest limitations. First, length and consistency. Clips are short, and the app does not always hold a character’s look perfectly from one generation to the next. Second, and this is the big one, cost. Video is the most expensive thing these apps generate, and it drains credit balances faster than anything else. I cover why that matters for your wallet in more detail alongside the broader move toward credit pricing.
Is it worth chasing yet
Depends entirely on what you want. If moving visuals are the specific reason you are looking, the feature has crossed the line from gimmick to genuinely worth trying. If you mainly want conversation, memory, or even just strong still images, video is not yet a good reason to pick one app over another. The image-first apps, like the one in my Candy AI review, still deliver better value per dollar for most people, because polished still generation is a solved problem and video is not.
How to try it without overspending
The same discipline that applies to all credit-based features applies double here:
- Use free credits first. Never buy a large pack to test video. Generate a few clips on the entry tier and judge the quality yourself.
- Watch the burn rate. Note how many credits a single clip costs versus an image. The gap is usually large.
- Judge consistency, not just one good clip. Anyone can cherry-pick a great result. Generate several and see how reliable it is.
Where this goes next
My honest expectation is that video quality keeps improving quickly and cost slowly comes down, which is the usual arc for generation features. Within a year it will likely be a standard offering rather than a headline one. For now, treat it as a bonus if your chosen app happens to do it well, not as the deciding factor. If you want a starting point that weighs visuals sensibly against everything else, my main ranking is where I keep it all in proportion.