🎬 How to make a video with AI
You can now go from a single sentence to a polished cinematic shot in under a minute. This guide explains exactly how — no prior experience needed.
1. Choose the right AI video model
Different models are built for different jobs. Some excel at photo-real human motion (Sora 2, Veo 3, Kling 2.5). Others are perfect for stylised animation, anime or product shots (Seedance, Pika, Runway). Start by browsing the Models page and look at the example videos — that tells you more than any spec sheet.
- Photoreal cinematic shots → Sora 2, Veo 3, Kling 2.5
- Fast & cheap drafts → Seedance, Pika
- Stylised / anime / 3D → Wan, Hailuo, MiniMax
2. Write a prompt that actually works
A great prompt describes the subject, the camera move, the lens and the lighting. Think like a director, not a search query. Even one line works if it's specific: 'Slow dolly-in on a wet neon-lit street, a woman in a red coat looking up, anamorphic 35mm, cinematic teal-and-orange grade.'
- Subject + action (what is happening)
- Camera move (dolly, orbit, crane, handheld)
- Lens & shot type (85mm, anamorphic 2.39:1, macro)
- Lighting & mood (golden hour, neon, volumetric haze)
3. Use the AI prompt booster
If writing a director-style prompt feels intimidating, just type a few words. Lumineer's prompt booster turns 'a woman walking in Paris at night' into a 120-word cinematic brief with camera, lens, grade and atmosphere — and if you attach a reference image, it analyses it to preserve your visual style.
4. Pick duration and resolution
Most video models charge by duration × resolution. Start with 5-second 720p clips while you experiment, then upscale your favourite to 1080p only when you're happy with the composition. It's cheaper and faster to iterate at the lowest tier.
5. Refine and re-roll
AI generation is non-deterministic — the same prompt produces a different shot every time. Generate 3 to 5 variations, pick the strongest, then tweak only one element (camera, lighting, or wardrobe) per iteration. That's how you converge on a great result instead of fighting the model.
Frequently asked
No. Everything runs in the browser — you only need an account and a few credits.
Most clips finish in 30 seconds to 3 minutes depending on the model and duration.
Yes — generations you make on Lumineer belong to you and can be used in client work, ads or social content.
Ready to put this into practice?
Open the studio and apply what you just learned in under a minute.
Make your first video