How do you avoid wasting credits on failed AI video generation cycles?

Spend video credits only on locked frames. Iterate cheap (still images, 720p drafts, single probe shots) until composition, character, product and camera direction are approved — then animate. Across documented productions this discipline keeps per-ad cost around $67–$150 even with 75–85% clip rejection baked in.

Start every shot as a still image, not a video. Generate the keyframe, iterate the framing, character and product on cheap image credits, lock it — only then spend video credits animating that exact locked frame. One documented production explicitly noted "I only spent video credits on locked frames," which is what holds per-ad cost near $125 even when ~85% of clips still get rejected during selection.

Run a single probe shot before any batch. Pick one full-frame shot containing your hardest variables — character, garment or product, set, skin, pose — generate just that one, and only scale up if it holds. For product work, validate at three focal distances (close-up, mid, wide) on the same product before committing the full run; if the product drifts across those three, no amount of iteration on a finished shot will fix it.

Ask for a pre-flight estimate. Before generating, prompt the invideo agent for a generation breakdown — how many reference sheets, how many video clips, how many UI screens it's planning — so you see the rough credit spend before you commit it. In one localization run this surfaced a ~570-credit ($145) estimate per ad up front, with no surprises after.

Drop the resolution for drafts. Seedance 2.0 at 720p renders meaningfully faster and cheaper than 1080p, and for vertical short-form it's visually sufficient — use 720p to test motion, camera move and timing, push to 1080p only on the locked take. Industry references on platforms like Runway peg the 720p→1080p jump at roughly a 40–60% credit increase, so this single toggle is one of the biggest credit levers you control.

Batch in fives, not fifteens. Generate video in batches of 5 clips, review, then continue — smaller batches let you catch a model going off early instead of paying for 15 clips before noticing. As one production put it: "smaller batches just give me more freedom to iterate quickly and early, and more importantly, they save a ton of credits compared to if I was generating all 15 clips at once."

Write the camera move in text and ask the invideo agent to play it back before generating. State framing, movement, speed and how the shot should land; have the agent confirm its planned execution; only then trigger the render. The agent will proactively ask the questions a good AD asks (duration, speed, landing) — answer them in chat instead of discovering the answer in a wasted 4-second clip.

Lock decisions in tiers, never skip ahead. The sequence that holds: brand context → characters → locations → keyframe per shot → video clips → voiceover. Each tier gets reviewed and locked before the next tier generates. Mid-project, ask the invideo agent for a "sitrep" — what's locked, what's still open across cast, wardrobe, location, music — so you never burn video credits on a shot whose creative direction isn't actually resolved.

Route each shot to the right model instead of brute-forcing one. invideo has all the current models available, and the invideo agent picks per task — Recraft or GPT-Image-2 / Nano Banana for stills depending on whether you need skin texture, lighting or product lock; Seedance 2.0 or Kling for video depending on the shot. When a shot looks "too AI," don't keep iterating on one model — render the same prompt across several and pick the best result. That's one comparison run, not twenty rerolls.

What this looks like in actuals: a 20-second product film at 300 credits ($75); a UGC ad at 500 credits ($125, including the 85% rejected clips); a full multi-platform brand campaign at 155 credits ($33); a 30–35-second real-estate B-roll package at 270 credits ($67); a full editorial campaign of 40 stills + 30 motion clips at 630 credits (~$150). Range across these: roughly $33–$150 per finished piece, with credit cost going up cleanly with how much motion and how much iteration the brief actually demands — not with how many things you re-rolled out of uncertainty.

Skip video entirely when a still will do. Internal review cuts, composition tests, lighting checks, animatic frames — none of these need video credits. Use images to make the decision, then spend video credits once on the locked frame.

Watch some of these to see what works for you:

I only spent video credits on locked frames.

— invideo creative team, documented UGC production