Keep the same motion excerpt, timing and framing while changing one approved character reference at a time. Define the review moments before generating the series. This lets you distinguish a character-specific problem from a change to the driving performance. Reusing one action does not guarantee that every body design will execute it equally well.
Freeze the performance source
Save a single approved motion file with a clear name. Record its excerpt, playback speed, crop and orientation. Do not create slightly different exports for each character unless that difference is intentional.
Write three review events, such as the first raised hand, the main turn and the return to rest. Use these same events throughout the set. Comparing only the first frame tests appearance, not whether the characters share a recognizable movement.
Keep the original recording available. If a later export loses preparation or introduces a speed change, you need a stable reference to identify it.
Make a character sheet that explains differences
| Character ID | Approved reference | Body coverage | Detail to protect |
|---|---|---|---|
| A | Named image file | Coverage needed by the phrase | Recognizable face and jacket |
| B | Named image file | Same reviewable coverage | Distinct hair and sleeve shape |
| C | Named image file | Any deliberate proportion difference | Separate silhouette and footwear |
These are planning fields, not invented inventory integration. Fill them with your actual original designs rather than borrowing the identity of a performer in the source.
Kling's official guide says its Motion Control Element Library supplies facial information, not clothing, hair or props. Consequently, evaluate those appearance elements separately instead of treating a face reference as the whole character specification.
Test contrasting designs first
Choose two characters with meaningfully different silhouettes or costumes. A narrow jacket and a loose sleeve can reveal different visibility problems during the same raised-arm gesture.
Keep the requested action and documented settings unchanged for this first comparison. If a tool exposes a seed or another repeatability setting, record it without assuming it guarantees identical motion.
Review action order, hand ownership and body balance, then review appearance. When one character fails, name the visible defect. Do not immediately rewrite the shared motion brief for the entire cast if the problem appears specific to that reference.
Separate shared revisions from individual repairs
A missing pause across both tests may justify revisiting the source excerpt or common settings. A hidden hand in one costume may instead justify a clearer reference for that character.
Retain version notes: what changed, which characters were tested and which review event improved or remained wrong. This is more useful than saving several unrelated files labelled “better.”
When the pair passes your chosen checks, expand the set gradually. Compare each new result with the same source, not only with the previous generated character. Otherwise, a drifting action can become the accidental reference for the next version.
Questions about consistency across a cast
Should all characters use identical body proportions?
No. Original designs can differ, but assess whether the intended movement remains readable for each. A successful human-shaped example does not validate every stylized body.
Can I change the costume while keeping the same motion test?
You can, but record it as a new appearance variable. Recheck hidden hands, sleeve boundaries and footwear before comparing timing.
Does this require automatic batch generation?
No. A small manually reviewed sequence can use the same controlled notes. Check actual product availability before planning around a batch feature or claiming that identities will remain stable without review.