Preview pass versus full render on a workflow

The mistakes that cost real credits in a multi-scene run are almost never visual. They are structural: the hook lands in scene three instead of scene one, the narration runs 40% long against the durations, two scenes say the same thing, the CTA is missing. None of those are things a renderer can get wrong, and all of them survive a beautiful render perfectly intact. You watch the finished cut, realise the order is wrong, and pay again.

A published recipe like HiveWell Raw Honey lists 310 credits at preview resolution and 620 at full. Run both passes, watch the finished cut, and only then notice the order is wrong, and the lesson cost 930 credits. Notice it while the scene list is still text and it cost nothing at all. The gap between those two outcomes is a habit, not a feature.

The two figures on a recipe, and what they mean

Every published workflow shows its credit cost as either one number or two separated by a slash. Where there are two, the first is the preview-resolution render and the second is the full-resolution one. It is the same recipe, the same scene list, the same model — rendered at two different resolutions with two different price tags.

The size of the gap varies more than people expect:

Recipe	Scenes	Preview	Full	Gap
HiveWell Raw Honey	10	310	620	2.0×
Lumee Glow Cat Glow-Up	28	1,092	2,156	~2.0×
Digestive System Explainer	11	396	495	1.25×
Morning Matcha Routine	3	405	405	none
Versely Podcast Promo	2	270	270	none

Two things fall out of that table. First, on most recipes the preview pass costs about half of the full one, which makes "validate cheap, commit once" a genuinely cheaper path than committing twice. Second, on some recipes there is no cheaper pass at all. Morning Matcha and Versely Podcast Promo publish a single figure. If you fork one of those, your first render is at full price, and every bit of validation you want has to happen before it.

That second case is the one people plan badly for, because they have internalised "there's always a draft" from the recipes where there is one.

Pass zero: the part that is text

Before any scene renders, a workflow exists as a list of scenes, and each scene is three pieces of writing plus a duration: an image prompt, a video prompt, and a voiceover line. That list is fully readable and fully editable before dispatch. This is where the structural review belongs, because at this stage everything is text and text does not cost credits to change.

What you can actually settle at this stage, with no render at all:

• Scene order and count. Read the voiceover lines top to bottom as a script. If it does not work as a script, it will not work as a video.
• The hook. Scene one either opens with something worth watching or it does not, and you can tell from the line.
• Narration pacing. Divide each scene's word count by its duration. Anything much above roughly two and a half words per second is going to be rushed, and no amount of rendering fixes a line that is too long for its shot.
• Redundancy. Two scenes carrying the same beat is the most common structural fault and the easiest to catch by reading.
• The last scene. Whether there is a CTA, and whether it is the one you wanted.

There is no "draft mode" toggle on a workflow — the only modes a saved workflow has are manual and scheduled, and neither of them is a cheaper render. This pass is just you reading the scene list before you press go, and it is the highest-leverage minute in the entire process, because a fix here costs an edit and a fix after the full render costs the full render.

Pass one: render at preview resolution

Once the structure reads correctly, the preview-resolution pass tells you the things text cannot: whether the character actually looks the same in scene two as in scene one, whether the motion in the prompts produces the motion you imagined, whether the cuts between scenes land or jar.

What to check on the preview render, in this order:

1. Continuity. Same face, same wardrobe, same palette, scene to scene. This is the failure that most reliably survives a good-looking single scene.
2. Motion. Did the video prompt produce movement, or a near-still frame with a drifting camera?
3. Cut points. Does each scene end somewhere a cut can live, or mid-gesture?
4. Audio against picture. Narration timing against what is on screen, not against the script.

What not to judge on a preview pass: fine detail, texture, small on-screen text, and grade. Those are resolution-dependent, and rejecting a recipe because the preview looks soft is how people end up paying for three full renders of a thing that was already fine. Resolution is the one variable the preview pass deliberately does not test, and it is worth understanding what moves when you turn it up — 4K instead of 1080p covers the shape of that.

Pass two: commit once

The full-resolution render should be a formality. If you find yourself running it twice, one of two things happened: you skipped pass zero and shipped a structural fault into an expensive render, or you used the preview pass to judge something it cannot tell you.

A discipline that holds up in practice: before you commit, write down what you expect the full render to change. If the honest answer is "nothing except sharpness," commit. If the answer includes anything about pacing, ordering, continuity, or script, go back a pass, because the full render will not fix those and you will be looking at the same problems at twice the price.

If the recipe is going to be run repeatedly rather than once, this is also the point to save it properly rather than treating it as a one-off. A workflow you have validated twice is worth keeping as a reusable recipe — the walkthrough is at save and reuse a video workflow — because the whole cost of the two-pass habit is amortised across every future run of the same format.

This is not the editor's 480p preview

Two different things share the word "preview" and conflating them causes real budgeting errors.

A workflow's preview-resolution render is a credit-charged render. It is cheaper than the full render on most recipes and identical in price on some, but it is a charge either way.

The editor's 480p preview is a different surface: setting preview: true on an edit gives a free 480p pass, and it carries a short per-user cooldown so it is not unlimited. The final export is charged once regardless of how many clips or features the timeline carries. That mechanic is specific to the editor's timeline, not to workflow runs, and the arithmetic is laid out in editor previews and final export.

So the two-pass habit inside a workflow and the iterate-then-export habit inside the editor are separate disciplines that happen to rhyme. Everything else on the platform costs credits, including both workflow passes — how credits work is the reference for what moves any given figure.

FAQ

Is the preview-resolution workflow render free?

No. Both published figures are credit charges. The preview pass is cheaper on most recipes, the same price on a few, and never zero. The only free render on the platform is the editor's 480p preview pass, which carries a short per-user cooldown.

How do I know whether a recipe even has a cheaper pass?

Look at the credit stat on its page in the workflow library. Two numbers separated by a slash means a preview pass exists at the lower figure. A single number means the recipe renders one way only, and your first render is at that price.

What should I check on a preview render that I cannot check in the script?

Continuity across scenes, whether the motion prompts actually produce motion, and whether scenes end at cuttable moments. Judge sharpness, texture and small on-screen text on the full render only, because those are the variables resolution controls.

If I fork a recipe and change scenes, do the two figures still apply?

Not directly. Both figures are for the recipe as published. Once you change the scene count, work from the per-scene rate instead, and remember that swapping the saved image-to-video model changes that rate for every scene at once.