AI Video Ad Generator for Product Ads
A product photo and a line of script become a person talking about it. Dialogue and picture come out of one sampling pass, so there is no separate voiceover to sync.
What CastTake is
In short: you give it a product photo and a line of script, and it returns a short video of a person talking about that product. The speech and the picture come out of one sampling pass, so the mouth matches the words without a separate voiceover step. A five-second ad costs 20 credits on the cheap model and 70 on the sharp one, where one credit is one US cent — so roughly 20 cents to test an angle and 70 cents to finish it.
How a product ad gets made here
- Bring a product photo. One clear picture of the thing you sell. It is optional — a written description works — but a photo is what keeps the product recognisably yours.
- Pick an angle. Unboxing, testimonial, problem and solution, product demo, or founder story. These are the five shapes nearly every UGC ad takes, and the one you pick changes the opening line more than any other choice.
- Let it write the prompt. Most people who need a product ad cannot write a video prompt, and being asked to is where they give up. The UGC ad generator writes it and leaves it on screen, editable — you can copy it and run it elsewhere.
- Generate, then test wide. Ten cheap variants to find the hook that works, then one expensive render of the winner.
What it costs
Pricing is per second of finished video, quoted before you press the button, and the number reserved is the number you were quoted. One credit is one US cent.
| Model | Resolution | Credits/sec | A 5-second ad |
|---|---|---|---|
| Grok Imagine 1.5 | 480p | 4 | 20 credits (~$0.20) |
| MiniMax H3 | 720p | 14 | 70 credits (~$0.70) |
| MiniMax H3 | 1080p | 22.4 | 112 credits (~$1.12) |
New accounts get 40 credits on the first day, which is one complete five-second ad rather than a fragment of one, and 20 more on each of the next four days when you come back and claim them. Full rates are on the models page and the plans are on pricing.
Which model to pick
Grok Imagine 1.5 for testing. It is three and a half times cheaper and the fastest to come back, and 480p is already past what most of a phone feed is watched at. Generate ten hooks here.
MiniMax H3 for the one you will spend money promoting. It renders at 720p and 1080p and accepts a reference image, which is how a presenter stays the same person across a campaign instead of drifting between batches.
The order matters more than the choice. Rendering everything at 1080p costs five and a half times as much and will not make a weak script work.
Frequently asked questions
- Do I need a product photo, or is a description enough?
- A description works and will produce a finished ad. A photo is what keeps the product recognisably yours rather than a plausible-looking substitute the model invented, so for anything you intend to actually run, use the photo.
- How long can one video be?
- Five, eight, ten or fifteen seconds. That is the range both models sell, and it is the range short-form ad slots use — a fifteen-second cap is a limit of this product, not a rounding of something longer.
- What happens to my credits if a generation fails?
- They come back. Credits are reserved when you submit and settled when the clip lands; a task that ends in failure or cancellation releases the reservation exactly once, so a generation that produced no video produces no charge. It is enforced in the ledger rather than handled case by case.
- What do I get without paying anything?
- Forty credits on your first day, which is one complete five-second ad rather than a fragment of one, and twenty more on each of the next four days when you come back and claim them. That is five finished clips on the cheap model without entering a card.
- Can I use the videos commercially?
- Yes — you keep the rights to what you make here, including the ads you run. What you cannot do is generate somebody real without their permission; likeness is the one place where owning the output does not settle the question.
- How is this different from generating a video and adding a voiceover?
- The speech and the picture come out of the same sampling pass, so the mouth was never separate from the words and there is no sync step to get wrong. The usual pipeline renders silent video, generates audio, then tries to reconcile them — which is where the uncanny half-second of drift comes from.