Render twice: a draft pass, then the delivery pass
Most shots that get thrown away are thrown away for reasons visible at any resolution: the wrong person spoke, the blocking broke, the line did not fit. Finding those at the cheapest tier and only then generating at the shipping tier changes what a series costs. As of 2026-09-12.
| Pass | Tier used | What is being checked | What a failure costs here |
|---|---|---|---|
| Draft pass | The cheapest tier the channel offers | Blocking, continuity, whether the dialogue fits the shot | One cheap regeneration |
| Delivery pass | The tier the episode ships at | Faces, hands, artefacts, legible text | A full-price regeneration |
| Fix pass | Delivery tier, selected shots only | Only the shots the delivery pass rejected | Full price, on a short list |
Inclusion rule. Passes over the same storyboard. A production that ships at the cheapest tier collapses the first two into one and is not a counter-example. Order. In the order the passes happen.
1Most failures are not resolution failures
Keep a note of why each discarded take was discarded and the list turns out to be dominated by staging: the line came from the wrong mouth, two characters swapped sides, a hand-off never happened, the shot ran out before the sentence did. None of that reads differently at a higher resolution.
The failures that genuinely need the shipping tier to detect are a shorter list: faces at distance, hands in motion, and anything with text in frame. Those are worth a full-price look, and they are the minority.
2The saving comes from the ratio, not the rate
A production that generates three takes per shot at the shipping tier pays three times the delivered duration at the dearest rate it uses. The same production with two of those takes moved to a cheap tier pays once at the dear rate and twice at the cheap one.
Nothing about the channel or the rate changed. The arithmetic changed because the takes moved, which is why the attempt ratio is worth more attention than the price list.
3Keep a take log, one line per attempt
Segment, tier, seconds, take number, and one sentence on what failed and what was changed. It takes a few seconds to write and it is the only record that shows which storyboard fields are repeatedly causing reshoots.
After one episode the log usually names two or three recurring causes. Fixing those in the template is worth more than any amount of prompt tinkering, because it removes the failures rather than retrying them.
4When the two passes are not worth separating
Where the shipping tier is already the cheapest one, there is nothing to save. The same is true for a production short enough that the overhead of two passes exceeds the generation it avoids.
Everywhere else the question is what fraction of takes are discarded. Below roughly one discard in ten the split earns little; above one in three it is the single largest lever in the budget.
Craft notes, not a register entry. No line here is attributed to a tool, and none should be read as a claim about one. The sourced material is on the register. Related: Four sheets, Storyboard fields.
5One pipeline's continuity
Serial work like “Two passes” depends on continuity; SceneMixer's guides describe how it keeps it:
SceneMixer turns a novel into a watchable short-drama series in one pipeline: upload the novel (Chinese or English) → AI extracts characters, scenes, and an episode-by-episode structure → generates consistent character and scene art → writes storyboard scripts → renders video segment by segment (16:9 by default, 9:16 in one setting) → composites full episodes.
SceneMixer, Turn a Novel into a Video Series with AI (read 2026-10-13)
Each segment is generated on whatever tier was selected at the time, and the episode composites them together.
SceneMixer, 480P, 720P or 1080P: Which Tier to Pick — “Can I mix tiers in one episode?” (read 2026-10-13)