AI Video Generator for Facebook Ads

What this tool does

• Ad Variants generates exactly 5 ad scripts in parallel, one per hook style: Curiosity, Problem–Solution, Social Proof, Urgency and Humor.
• The script step calls VdoBloom's ad-script endpoint, which performs no credit deduction — credits are charged only when you render a chosen variant as video.
• Each variant targets a 15-second ad and is capped at 3,000 characters so it fits the prompt box of the model you'll render it on.
• Every generated script is a timestamped shot list: per shot it specifies visual description, camera movement, lighting style, motion details, the transition out, and the voiceover line, with the timeline summing to the target duration.
• The final shot is mandatory in every script — product visible, brand present, tagline integrated — because the director prompt forbids ending abruptly.
• The script engine is instructed to upgrade a weak reference photo rather than copy it: cluttered backgrounds, poor lighting and unprofessional packaging in your upload are not carried into the ad description.
• Meta's ad specs use 9:16 for Reels and Stories placements, 4:5 for feed and 1:1 for square. In VdoBloom's video roster, 4:5 is available on HappyHorse 1.1 Text-to-Video and Reference-to-Video — and on the Ad Variants tab, which runs an image-to-video picker, Reference-to-Video is the one you can reach; 9:16 and 1:1 are available across Runway, Kling, Seedance, Wan 2.7, PixVerse, Vidu, MiniMax H3, Grok Imagine and P-Video.
• For static 4:5 feed creative, the image roster includes 4:5 on Nano Banana Pro, Nano Banana 2 and 2 Lite, GPT Image 2 (text-to-image and image-to-image), Riverflow 2.0 Pro and Fast, and ImagineArt 2.0.
• Ad Variants defaults to Seedance 2 at 9:16, 10 seconds and 720p, and prices image-to-video rather than text-to-video because the flow always starts from your product image.
• Real render costs from the credit table: Wan 2.7 image-to-video 720p 10s = 64 credits; Seedance 2 image-to-video 480p 10s = 76 credits and 720p 10s = 163; Kling V3 Turbo bills 7.2 credits per second at 720p.

How it works

1. 1.Describe the product and upload one image

   Enter the product name, the single key benefit, the target audience and optionally a brand name, then upload the product or influencer photo the video will animate. The image is required before rendering because Ad Variants runs an image-to-video pipeline.

2. 2.Generate all five hooks at once

   The tool sends five parallel script requests, one per hook style, and returns each as a titled, timestamped shot list with a voiceover track. Nothing is deducted for this step, so you can regenerate the set if none of the angles land.

3. 3.Read the scripts and pick the angle

   Expand each variant to read the full script. Judge them on the opening shot and the voiceover's first line — that's the part Facebook shows before the viewer decides to keep watching.

4. 4.Match the format to the placement

   Pick 9:16 for Reels and Stories placements, 1:1 for square feed. If you specifically need Meta's 4:5 feed ratio in video, switch the model to HappyHorse 1.1 Reference-to-Video — it is the only variant exposing 4:5 and 5:4 that this tab's image-to-video picker carries (HappyHorse 1.1 Text-to-Video also has it, from the Text-to-Video tab).

5. 5.Render, then re-render the runner-up

   Generate the winning variant, then come back and render the second-place hook on the same product image. Two videos from one product photo is the cheapest form of creative testing available in the tool.

Frequently asked questions

Related tools