It started as a complaint about re-rendering.
NeoMotion is built by Adi Suandharu, a product engineer from Indonesia. Before this there were two other AI products, and before those, nine years of shipping software for large organisations — long enough to develop a specific allergy to tools that demo beautifully and then fall apart the moment real work touches them.
The idea came out of making explainer videos the slow way. The AI part was genuinely good: it could write, it could plan a storyboard, it could generate imagery. Then it handed over a finished video, and every small fix — a wrong statistic, a brand colour, a title that read badly at the last second — meant starting the generation over. Nothing was wrong with the intelligence. Everything was wrong with the output format.
So the product got built backwards from the end of the pipeline instead of the start. Not “what can the model generate” but “what does a motion designer need to still be able to edit” — native layers, keyframes, real presets, a real timeline. Then the agent was pointed at that, given hands, and told to do the production rather than describe it.
That turned out to be the harder half. Any model can write a script. Almost none can go and find the b-roll, cut the subject out of it, cast a preset that fits the message, and place it on a timeline you can still take apart. That gap is the whole product.
It is still early. The installers are not code-signed, parts of the roadmap are openly unbuilt, and the first users are shaping what comes next in a way that will never be as true again as it is right now. If that is interesting rather than alarming, there is a founding cohort with your name on it.