How to Generate a Product Launch Video with Moonvalley AI
Moonvalley generates short cinematic video clips from text prompts. For a 30-second product teaser, the working method is: build six 5-second shots from one visual style prompt, then cut them together with a real product still or screen recording inserted at the reveal moment.
Pure AI launch videos look like AI. Hybrid ones do not.
What kind of video does Moonvalley produce?
The current product, Marey, generates text-to-video clips with a cinematic default style — shallow depth of field, filmic color, motion blur. Clip length is short, typically in the range of a few seconds per generation, which suits editing rather than long-form.
It has shipped in tiers aimed at studios and brands rather than casual users, with pricing structured for teams. That matters: this is not a $10-a-month consumer toy, and you should confirm current plan details before budgeting.
How do you plan a 30-second launch teaser?
Write the shot list before touching the tool. Six shots, five seconds each, is a clean structure for a 30-second cut.
- Establish — wide atmospheric shot, product absent.
- Approach — camera moves in, tension builds.
- Context — hands, a desk, a room; human scale.
- Reveal — the product itself. Use a real photo or render here.
- Detail — close-up of one feature, abstract is fine.
- Close — logo card with a slow push.
Shots 1, 2, 3, and 5 can come from the model. Shots 4 and 6 should not — the product must be accurate and the logo must be crisp.
How do you keep six shots looking like one film?
Consistency is the hard part of AI video, harder than getting a single good clip.
Anchor your prompts with a fixed style block repeated verbatim across all six generations. Something like: “35mm film grain, warm amber key light, shallow depth of field, slow dolly, no text overlay.”
Repeat it every time. Do not paraphrase. Small wording changes shift the color grade enough that the shots will not cut together.
Second trick: generate each shot six to ten times and pick the best. The accept rate for a usable 5-second clip in my experience sits around one in five for complex motion, and closer to one in three for simple camera moves. Budget your generation credits accordingly.
Does AI video handle the product accurately?
No — and this is the single biggest trap.
Text-to-video models hallucinate details. Buttons move, text on packaging becomes gibberish, a phone’s camera layout shifts between shots. For a launch video, that is fatal. Viewers notice.
Fix it with compositing. Generate the environment shots with AI, shoot or render the product separately, and composite in the product using DaVinci Resolve, Premiere, or After Effects. Screen replacements take 10 minutes per shot once the background plate exists.
If you must generate the product with AI, keep it small in frame and moving. Detail scrutiny comes from stillness and close-ups.
What are the practical limits?
| Limit | What it means |
|---|---|
| Clip length | A few seconds per generation; long takes must be stitched |
| Control | Prompt-level only; no keyframe or motion-path input |
| Consistency | Requires a repeated style block and manual selection |
| Text rendering | Unreliable; add typography in the edit |
| Cost | Team-tier pricing; check current plans before committing |
Budget expectations should be set against alternatives. A stock footage license for a comparable 30-second spot runs a few hundred dollars and gives you real products. AI video gives you invented ones faster.
Who is this for?
Useful for agencies pitching concept boards before production money is approved. Useful for brands that need eight regional variations of a mood piece. Useful for pre-visualizing a spot before a real shoot.
Not useful for: shooting the actual product hero shot, anything requiring legal review of what appears on screen, or anyone needing reproducible frame-accurate results across versions.
Quick Summary
- Moonvalley generates cinematic clips of a few seconds; a 30-second teaser means stitching six of them.
- Repeat the exact same style sentence in every prompt or the shots will not cut together.
- Expect roughly one usable clip per five generations for complex motion — budget credits for retries.
- Never generate the product itself; composite a real photo or render at the reveal.
- Team-tier pricing; confirm current plan details before committing a budget.
