If you've spent any time looking for an AI video generator online, you've already met the pattern. Beautiful homepage, eight-second hero clip showing a dragon flying through clouds, "start free" button. You click it. You get a fourteen-day trial. The trial requires a credit card. The "free credits" run out after two renders. The output is half the resolution of the marketing reel.
We built ours different because we got tired of that. This post is about what an AI video generator actually does today, where the lines are, and how to pick the right one (ours or otherwise).
Every AI video generator on the market right now is one of two things, sometimes both. Text-to-video (T2V) takes a sentence — "a cinematic shot of a Bedouin walking across red dunes at sunset" — and renders motion. Image-to-video (I2V) takes a still image you already have and gives it motion: the woman in your headshot turns her head, the product on your table rotates, the city skyline gets a slow camera pan.
T2V is the showy one. It's what gets shared on Twitter. I2V is the one that actually gets used in production, because the quality bar is way higher when you're starting from a real image and the motion is the only AI part.
All free at the tool level. Live here.
It means the rendering happens on hardware we own, the model weights are open-source or self-hosted, and we don't pay per-token API fees that we'd have to pass on to you. The catch — and there's always a catch — is queue time. During peak hours you might wait three minutes. During off-hours it's under thirty seconds. We don't gate by credits because we don't pay per render.
The math on a paid competitor: $25/month for 200 credits. Each video costs 5–20 credits. So 10–40 videos a month for $25. Or: unlimited videos here for an email.
If you respect those edges, you'll get usable output every time. If you try to use AI video for something it can't do yet, you'll spend forty minutes troubleshooting and end up with garbage. That's not a tool problem. That's a model maturity problem, and it applies to every AI video generator on the planet right now.
Start with the subject, then add framing, then lighting, then motion. Example: "Bedouin in indigo robes, medium shot, golden hour backlight, slow dolly-in." Each comma adds a level of control. The model handles all four directives.
For I2V, pick an image where the subject is centered, the background is simple, and the motion you want is obvious from context. A photo of a coffee cup with steam rising will get you a great "steam swirling" loop. A photo of a person mid-stride works for "walking forward."
Video is one piece of a larger media engine. The same brain that powers our cartoon generator, our music generator, and our lipsync tool drives the video pipeline. When the desktop app ships, all of this becomes one local install — no queue, no upload, runs on your GPU.
Next month: 4K I2V output, voice-to-video (speak the description, render the clip), and a one-shot "script-to-film" mode where you paste a written scene and get a fully stitched short back. That last one is the unlock for indie filmmakers who've been priced out of the production pipeline.
Free renders now. Desktop install when QADIR OS ships. No credit card, ever.
Open the Video Generator