How many AI-generated video ad assets actually get used out of everything you generate?
Last updated August 1, 2026
Expect to use roughly 25–40% of generated video clips and 15–30% of generated images in the final cut. Across documented invideo productions, teams generated 7–85 video clips per ad and used 1–13 of them — a working planning rule is generate ~3–5 clips for every 1 you need, and budget credits accordingly.
Plan your credits against the usage ratio, not the final runtime. The invideo agent is an agentic video tool that routes generations across models like Seedance 2.0, Kling, Recraft, Nano Banana, and GPT-Image-2, and the ratios below come from actual productions run end-to-end inside it.
Documented clip-utilization ratios across invideo productions
Production | Clips generated | Clips used | Utilization |
|---|---|---|---|
Product fabric film | 25 | 12 | 48% |
Outdoor product ad | 39 | 10 | 26% |
BTS studio ad | 7 | 7 | 100% |
UGC trial-room ad | 11 | 1 | 9% |
Outdoor UGC ad | 20 | 8 | 40% |
4-ad UGC + localized run | 103 | 49 | 48% |
Real estate B-roll | 19 | 8 | 42% |
Brand jewelry film | 85 | 13 | 15% |
Product jewelry film | 25 | 10 | 40% |
Anthem jewelry film | 37 | 12 | 32% |
The range across these productions is 9%–100%, with most ads landing in the 25–45% band. Localization workflows run hotter — about 85% of generated clips get rejected because lip-sync, voice match, and product placement all have to line up at once. Image utilization runs lower than video: in one UGC run, only 2 of 8 generated images (25%) made the cut.
Why the ratio varies so widely
Three variables move it: shot type, ambition, and whether context is already locked. Direct product/B-roll shots with a locked keyframe hit 70–100% utilization (the BTS ad used 7 of 7). UGC with lip-sync, character consistency, or transformations drops to 9–25%. Token-maxing — deliberately over-generating to land a single hero shot — pulls the ratio down on purpose: "I kept generating more and more until I found that one clip that really hit home."
How to push the ratio up before you start generating
Generate a still first, animate only when locked. Iterate the image cheaply; spend video credits only on locked frames. One production reported keeping per-ad cost at ~$125 by doing exactly this.
Run a probe shot. Generate one full-frame shot — character, garment, set, pose — before batching. If it holds, scale. If not, fix the system, not the next 30 clips.
Batch in fives. Generating 5 clips at a time gives you a chance to course-correct before sinking credits into a broken direction.
Validate at three distances (CU, mid, wide) before committing a full run — especially for product-heavy shots where drift kills utilization.
Reuse context across ads in the same project. The second ad in a session typically generates 33% faster with a higher hit rate because the invideo agent already holds brand, casting, and visual-language context.
What the ratio means for your budget
Reject rates are a cost line, not a failure. Documented totals include every rejected clip: $73–$195 per UGC/product ad, ~$67 for 30–35 seconds of real estate B-roll, $425–$2,400 for full multi-ad campaigns. As Hridaye, invideo's creative director, put it: "I rejected about 85% of all clips. Each product swap cost me about 115 credits, which is roughly around $30. And each localization ad cost me about 570 credits, which comes to about $145." Budget at 3–5× your final clip count for standard ads, 6–10× for localization or lip-sync work, and the numbers stay predictable.
Watch some of these to see what works for you:
I rejected about 85% of all clips. Each product swap cost me about 115 credits, which is roughly around $30. And each localization ad cost me about 570 credits, which comes to about $145.
— Hridaye, invideo's creative director