Read a workflow's credit estimate before you run it
A 3-scene recipe can cost more than a 15-scene one. What the two published credit figures include, what they leave out, and how to approve a run on cost.
The NYC Street Interview recipe is three scenes long, thirty seconds of finished video, and lists a flat 504 credits — one figure, no cheaper option. The AlignPro Posture Corrector recipe is fifteen scenes, nearly twice as long on screen, and starts at 451. The eleven-scene Digestive System Explainer lists 396 at preview resolution and 495 at full, which puts it under the three-scene recipe at either end.
If your mental model is "more scenes costs more," those three break it, and they break it in the direction that matters: the cheap-looking short recipe is the expensive one. Every published workflow carries its scene count and its credit estimate on the same page, and reading the two together is the difference between approving a run you understand and finding out afterwards.
The numbers a workflow page actually publishes
Open any recipe in the workflow library and the recipe block gives you the same set of stats every time:
- Runtime — the length of the finished cut, summed from each scene's duration.
- Scenes — how many discrete shots get generated and stitched.
- Aspect ratio — 9:16 or 16:9, which tells you where it is meant to be posted.
- Credit cost — one figure, or two separated by a slash.
- Video model — the image-to-video model every scene is animated with.
- Category, published date, and, where it applies, how many creators have cloned it.
The credit figure is the one worth slowing down on, because it is the only stat on the page that is a decision rather than a description.
What the estimate is counting
A scene workflow is not one generation. Each scene runs as a pair: a still image is generated first, then that still is animated by the workflow's saved image-to-video model, and the finished clips are stitched into one video at the end. So a fifteen-scene recipe is thirty generations plus a stitch, and the published figure is the total of all of it.
That is why the model line matters more than the scene count. The image half is comparatively stable across recipes; the video half is where the money goes, and image-to-video models sit at very different credit levels. Divide the total by the scene count and the pattern falls out immediately:
| Workflow | Scenes | Credits (preview / full) | Per scene | Video model |
|---|---|---|---|---|
| Viral Panel Reaction Reel | 4 | 60 / 120 | 15 | Seedance v1.5 Pro Image to Video |
| HiveWell Raw Honey | 10 | 310 / 620 | 31 | Vidu Q3 Image to Video |
| Digestive System Explainer | 11 | 396 / 495 | 36 | Kling 3 Turbo Image to Video |
| Lumee Glow Cat Glow-Up | 28 | 1,092 / 2,156 | 39 | Vidu Q3 Image to Video |
| Primo Protein vs XXX | 11 | 1,100 / 2,200 | 100 | VEO 3.1 Fast I2V |
| NYC Street Interview | 3 | 504 | 168 | Kling Video V3 Pro Image to Video |
Read the per-scene column and the earlier oddity stops being odd. Lumee is twenty-eight scenes and still lands below Primo's eleven, because a Vidu Q3 scene and a VEO 3.1 Fast scene are not comparable units. A scene on a premium model can cost several times what a scene on a cheaper one costs, which is why a three-scene recipe on Kling Video V3 Pro carries a much heavier per-scene commitment than a fifteen-scene recipe on Vidu Q3.
Per scene is also the number to carry across recipes when you are budgeting. Total credits tell you what this particular cut costs. Credits per scene tell you what your cut will cost when you fork the recipe and change the scene count, which is what people actually do.
Why some recipes list two figures and some list one
Where a workflow shows two numbers, the first is the preview-resolution render and the second is the full-resolution one. Where it shows a single number, the two are identical in the catalog and there is no cheaper pass to take.
The gap is not constant. Most two-figure recipes roughly double from preview to full: HiveWell goes 310 to 620, Primo goes 1,100 to 2,200, Lumee goes 1,092 to 2,156. The Digestive System Explainer is the outlier in the current library, moving only from 396 to 495. And three recipes, all on Kling Video V3 Pro Image to Video, publish one figure only.
So "run the cheap pass first" is a habit that pays very differently depending on which recipe you picked, and on one of them it does not exist at all. Check before you plan around it. If you want the general version of that discipline applied to the timeline rather than the recipe, the editor previews and final export breakdown covers where the charge actually lands.
What the estimate does not include
This is where approvals go wrong. The published figure covers the scenes and the stitch. It does not automatically cover everything you will end up wanting on the finished video.
- Captions. Of the seventeen published recipes, exactly one carries a caption style on the recipe itself. For everything else, adding captions is a separate step on its own pricing, and a recipe tagged with a caption look in its tag list is not the same as a recipe that renders captions as part of the run.
- Dubbing, upscaling, reframing. All post-run operations, all priced on their own.
- Rerolls. The estimate prices one clean pass through the scene list. A recipe where you redo three scenes is not the published number.
- Your edits. Fork a recipe and add four scenes and the published total stops applying. The per-scene figure is what survives the fork.
The figures are each recipe's own published cost and they move when the catalog moves. Treat a number you screenshotted last quarter as a rough guide, not a quote.
Approving or rejecting a run on cost alone
You can make this call in under a minute without opening the scene list, and most of the time you should.
- Read the total and your balance together. If the total is a meaningful fraction of what you have left, the interesting question is not "can I afford this" but "is this the run I want to spend it on."
- Divide by scenes. Anything above roughly a hundred credits a scene in the current library is a premium-model recipe. That is a legitimate choice for a hero asset and a poor one for a first test.
- Check whether a second figure exists. If it does, the preview pass is your validation budget. If it does not, your first render is your only render at that price.
- Add the finishing steps you know you will want. Captions, a dub, an upscale. If those push it past your ceiling, the recipe is out regardless of how good the preview clip looks.
- Ask for the estimate against your actual balance if the run is large or multi-step. The agent will total a planned job using the same pricing logic that does the real charging and show your balance next to it — the walkthrough is at check my credits before generating. Estimating does not itself charge credits.
The failure mode this prevents is not overspending in one dramatic run. It is the slow version: approving four recipes over a week without ever converting a total into a per-scene rate, then being surprised by where the balance went. For the general mechanics behind any credit figure, how credits work covers what moves a price.
FAQ
Why does a shorter workflow sometimes cost more than a longer one?
Because the video model is the dominant variable, not the scene count. The three-scene NYC Street Interview lists a flat 504 credits; the eleven-scene Digestive System Explainer lists 396 at preview resolution and 495 at full. Always divide the total by the scene count before comparing two recipes.
What does the estimate include, and what is extra?
The published figure covers generating every scene and stitching them into one final video. Post-run finishing work — captions, dubbing, upscaling, reframing — is priced separately and is not folded into the recipe's number.
If I fork a recipe and change the scene count, what happens to the estimate?
The published total stops applying and the per-scene rate becomes your planning number. Multiply it by your new scene count, then re-check against the model, because changing the saved image-to-video model changes the per-scene rate too.
Is the preview figure a free pass?
No. Both figures are credit costs; the preview-resolution render is cheaper, not free. Nothing in a workflow run is free — the only free render on the platform is the editor's 480p preview pass, which carries a short per-user cooldown and is a different surface entirely. When two recipes both look right, compare their per-scene rates against what you actually get for them, which is the frame the quality-per-credit reporting uses.