A/B Cover Stills From One Locked Storyboard Body
Derive 2–3 cover stills from locked hook/proof frames to test CTR without regenerating the Reel body—plus mismatch and bait rules.
A/B Cover Stills From One Locked Storyboard Body
Test thumbnail and cover variants without regenerating the Reel body. CTR experiments should change the door—not rebuild the house.
Creators optimizing click-through while keeping teach content stable already know hook A/B on a locked body (A/B hooks on one locked storyboard body). Cover-must-match-hook checklists cover alignment (thumbnails that match the hook). This piece is deriving cover candidates from locked board stills and logging fair tests.
Derive 2–3 covers from approved frames
Pull candidates only from the locked board:
| Variant | Source frame | Good for | Avoid |
|---|---|---|---|
| A — Hook still | Beat 1 approved still | Promise clarity | Adding new claims |
| B — Proof still | Mid-Reel proof crop | Credibility seekers | Spoiling the whole method |
| C — Framework still | Named checklist/formula card | Save-oriented audiences | Tiny unreadable type |
Do not invent a fourth “mystery face” cover that never appears in the Reel. Pull cover candidates from locked board stills only—then crop for platform chrome.
Mismatch rules vs bait
Allowed mismatch (mild)
- Tighter crop than the in-Reel framing
- Higher caption contrast than motion frames
- Platform-safe margin adjustments
Bait (forbidden)
- Numbers or outcomes not in the Reel
- Faces/products that never appear
- “Part 2” energy when the asset is standalone
- Sexualized or shock crops unrelated to teach intent
- Before/after that implies a result you do not show
If the viewer feels tricked in the first three seconds, you did not win a test—you taxed trust. Cover composition craft still matters (AI cover stills that survive crops).
Log template
Run covers as controlled swaps:
- Reel ID / board version (body locked)
- Variant — A/B/C with source beat ID
- Platform + placement
- Start / end time
- Primary metric — CTR or open rate you already use
- Secondary — 3s holds, saves (watch for bait spikes with fast exits)
- Winner rule — predeclared sample window
- Notes — crop/safe-zone issues
Do not change captions, hooks, and covers in the same window. One variable per test. First-frame clarity inside the video remains independent (half-second grasp).
Production sequence
- Lock storyboard body + hook
- Export stills for beats tagged
COVER-CANDIDATE - Design 2–3 crops; approve against bait rules
- Publish with variant A; rotate B/C on schedule
- Promote winner to default; archive losers with reason
- Only then consider a new hook test—not a new body
Approve stills before motion elsewhere in the pipeline (approve stills before motion); covers inherit that discipline.
Cropping for platform chrome
Each platform eats different corners. When deriving covers:
- Keep primary subject and title in the center-safe rectangle
- Avoid locking meaning into the top 15% or bottom 20%
- Test with UI overlay templates before declaring a winner
- Export master 1080×1920 (or your pipeline master) plus platform-specific crops if needed
A variant that wins on one platform can lose on another because of chrome—not because the teach changed. Log platform separately.
Statistical humility
Predeclare windows and avoid peeking every hour to “call” a winner. Small samples lie. If traffic is thin, extend the window rather than shipping bait. Secondary metrics (rapid exits after click) catch covers that overpromise.
When to break the locked body
Only reopen the body if:
- All covers underperform and watch-through is weak (hook/body problem)
- Product UI in proof covers is obsolete
- Claims were revised and stills show old numbers
Otherwise keep experimenting on the door. Body churn destroys test hygiene and budget.
Archive hygiene
Store losers with:
- Board version
- Crop file
- Metric snapshot
- Reason code (
low_ctr,bait_risk,unreadable_type)
Future you will thank present you when a “new” cover idea is last quarter’s loser.
Creative boundaries for designers
Brief designers (or models) with:
- “Source beat IDs only—no new scenes”
- Hex/type from style lock
- Bait forbidden list
- Safe-zone overlay file
This keeps cover craft sharp without reopening the storyboard body for decoration.
Soft production tip
Pull cover candidates from locked board stills only. Storyboard-first workflows such as Scro keep those frames versioned with the body, so CTR tests stay honest and cheap—no full regenerations for a thumbnail whim.
Checklist
- Body locked and versioned
- 2–3 covers mapped to real beats
- Bait rules checked
- One variable per test window
- Log filled before launch
- Winner rule predeclared
- Losers archived with notes
Covers open the door; the locked storyboard keeps the teach trustworthy after the click.