The 20x gap that shapes the whole workflow
A hook-testing still costs 12 credits at 720p and 14 at 1080p. A 5-second clip costs 300. That roughly 20x gap is the entire strategy: stills are for finding out what works, video is for scaling what already did. Even the expensive image lane keeps the shape of the argument — a UGC product ad render, where you attach your own product photo and the engine reproduces that exact item, costs 56 credits, still more than five times cheaper than animating a frame. Teams that animate everything burn their allowance on unproven creative; teams that animate winners get video performance at still-testing prices.
The stills-to-video academy guide covers the product mechanics — this post is the ad workflow wrapped around them, from cheap testing to shipped clip.
Step 1 — Find winners with still testing
Run your creative testing at the image layer first — and run it on the standard photoshoot lane, because a hook, a setting and a framing can all be judged without your real packaging in shot. Twenty to forty variants at 1080p, 14 credits each, is 280–560 credits for a full testing round — the price of one or two clips. Test in real placements, and let spend-weighted results, not taste, pick the frames that graduate to motion.
The AI UGC ads use case covers still-side production in detail; the short version is one trained persona, one preset, and variation confined to the variables you are actually testing. When a hook wins and the finished ad has to show your real product, re-render that one frame as a UGC product ad — pick the character, attach a single product photo, pick the beat — and 56 credits buys a frame with the correct packaging already in it, ready to animate.
Resist the urge to skip this step because you 'already know' which creative works. The entire reason this workflow is cheap is that stills fail at 14 credits instead of 300 — every skipped testing round is a bet of more than twenty times the stake on your own taste.
Step 2 — Pick frames that will survive being animated
Not every winning still is a good clip candidate. A frame that converts as a static image can still fight the animation — too tight, too busy, or dependent on overlaid text that motion will smear. Before spending 210 credits, check the frame against four criteria:
- Clear subject-product relationship: the product is visible and held or used naturally
- Room to move: compositions with a little negative space animate better than tight crops
- Readable emotion: the expression carries the hook even before motion is added
- No text baked into the frame: overlay captions in your editor, not in the render
Step 3 — Write motion prompts that respect the identity
Subtle, physical, single-intention motion works best: 'she turns toward the camera and smiles', 'slow push-in while she lifts the bottle'. Stacked actions and dramatic camera moves fight the source frame and produce exactly the artifacts viewers notice first.
One motion idea per clip is the rule. If two ideas both matter, that is two clips — and at 210 credits each, two focused clips that work beat one ambitious clip that almost does.
Write the prompt against the frame you have, not the ad you imagine: reference what is visibly in the still — her pose, the product's position, the light source — and the motion inherits believability from the image instead of fighting it.
Step 4 — Generate 5 seconds first, extend winners to 10
The 5-second image-to-video clip at 210 credits is your video testing unit. Ship it into placements with captions and sound, and let performance decide which clips earn the 10-second, 420-credit treatment — usually structured as the proven hook up front with a demonstration or payoff in the back half.
Doubling length before the short version proves out just doubles the cost of being wrong. The 10-second budget belongs to clips that have already held attention for five.
Step 5 — Cut, caption, label and ship
Treat the generated clip as raw footage, not a finished ad. Caption it — most feed viewing happens muted — front-load the hook into the first second, and apply the platform's AI-content label: realistic synthetic people need labeling on every major platform, and unlabeled synthetic media risks distribution penalties long before legal ones.
Keep the persona presented as an actor in branded creative, never as a real customer reporting personal results — that line is a policy, platform and legal requirement all at once.
Archive the winning combinations — source still, motion prompt, final cut — in a shared document per product. Winning motion patterns transfer across products far more reliably than winning stills do, and the library turns next month's testing into iteration instead of exploration.
Common failure modes (and the cheap fixes)
Most wasted video credits trace back to a handful of repeat mistakes, all of them avoidable before generation starts:
- Animating a crop: faces near the frame edge distort first — regenerate the still with more headroom before animating
- Two motions in one prompt: 'she waves and walks away' splits the clip's attention; pick one
- Chasing a broken clip with re-rolls: if two attempts fail the same way, the source frame is the problem, not the seed
- Skipping the sound pass: a silent clip shipped to a sound-on platform underperforms for reasons that have nothing to do with the visuals
What a month of this costs
A steady pipeline — four still-testing rounds on the standard lane (about forty images, 560 credits), eight winners re-rendered as product ads (8 × 56 = 448 credits), eight 5-second clips (1,680 credits) and two 10-second extensions (840 credits) — lands around 3,500 credits a month. That is Creator plan territory at €69/month for 9,000 credits, with headroom for retries; teams running the same motion per client scale the identical math on Studio or Agency.
The number worth watching is not total spend but the ratio of credits spent on proven versus speculative creative. Only the 560 credits of hook testing is speculative here, so the workflow above keeps roughly seven of every eight credits on frames that already earned their way in.
