Agency Handoff: A Freelancer-Ready Storyboard Pack Spec
Brief freelancers so AI explainer work doesn’t drift: pack contents (intent, beat table, style lock, stills, do/don’t, QA gate) plus a sample one-pager.
Agency Handoff: A Freelancer-Ready Storyboard Pack Spec
Brief freelancers and agencies with a storyboard pack—intent, beat table, style lock, stills, do/don’t, QA gate—so AI explainer work does not drift from your teach claim.
Marketing leads outsourcing teach-video production lose weeks when the brief is a moodboard and a Google Doc novel. The pack is the approval artifact before any vendor renders clips. Build on how marketers brief an AI storyboard and annotation for useful feedback.
Pack contents (minimum)
| Section | Must include | Why |
|---|---|---|
| Intent one-liner | Audience + outcome claim | Prevents pretty off-topic films |
| Beat table | Time, job, visual type, on-screen words | Shared spine |
| Style lock | Character, palette, type rules, set | Stops look drift |
| Reference stills | Hook + mechanism + end card | Phone-size truth |
| Do / Don’t | 5–10 bullets | Cheap clarity |
| QA gate | Checklist + sign-off roles | Defines "done" |
| Rights / claims | What may not be said | Legal/trust |
| Delivery specs | Ratios, length, captions, file naming | Ops sanity |
If a section is missing, freelancers will invent it—usually wrongly.
Sample one-pager (fill-in)
Working title: Viewer vs Admin invites Intent: Teach invite → Role → Viewer so audits stop failing Audience: Admins who already use the product weekly Length / ratios: 40–45s; master 9:16 (+ 1:1 notes) Host: Character ID "Alex-v3"; wardrobe code NAVY-TEAL UI version: 2026-10-28 Billing/Invite dialog
Beat table (abbrev):
| t | Job | Visual | On-screen |
|---|---|---|---|
| 0–3 | Hook | Card+host | "Viewer invite ≠ Admin invite" |
| 3–10 | Problem | UI crop | "Everyone Admin → audit mess" |
| 10–28 | Mechanism | UI | "Invite → Role → Viewer" |
| 28–38 | Proof | UI | Confirmation + role badge |
| 38–45 | Soft CTA | End card | "Save the invite path" |
Do: crop to role picker; mute captions; match end card template v2 Don’t: invent Admin upsell; change host face; stretch 9:16 to 16:9 without notes
QA: run post-render checklist vs this pack; PMM signs claims; producer signs mute pass
Attach stills. The one-pager is the cover; stills are the contract.
How to run the handoff
- Internal approve the pack (intent + stills) before vendor kickoff
- Walk through beat jobs on a 15-minute call—not a one-hour taste debate
- Vendor returns updated board marks, not finished renders first
- Approve board diff
- Render → QA vs pack → publish
Skipping step 3 is how you pay for three wrong full renders. Review rituals keep comments on beat IDs.
What “done” means
Done means: every QA gate item passes against the pack, claims match, ratios match delivery specs, and line/host locks match the style section. Done does not mean "the freelancer likes it" or "it feels premium."
For batch weeks, hand one style lock plus N beat packs—see batch from one style lock.
File tree freelancers actually use
/pack-invites-viewer-role/
00-one-pager.md
01-beat-table.csv
02-style-lock.md
03-stills/hook.png
03-stills/mechanism.png
03-stills/endcard.png
04-do-dont.md
05-qa-gate.md
06-claims-rights.mdZip it. Version it (pack-v1.2). When the product UI changes, ship pack-v1.3 with a changelog of which files moved—not a Slack thread of screenshots.
Payment and revision policy (brief note)
Define in the pack: how many board revision rounds are included before render, and that renders are out of scope until board approval. This protects both sides and keeps AI credit spend tied to approved beats.
Scorecard for vendor drafts
When the freelancer returns a board or first cut, score:
| Criterion | 0–2 |
|---|---|
| Intent match | |
| Beat jobs intact | |
| On-screen words match pack | |
| Style lock obeyed | |
| Safe zones / mute captions | |
| CTA matches soft ask |
A cut can look expensive and still score low. Low scores go back to board—not to "make it prettier."
Internal RACI
| Task | Marketing lead | PMM | Freelancer |
|---|---|---|---|
| Pack v1 | A | C | I |
| Board revisions | C | C | R |
| Claim sign-off | C | A | I |
| Render | I | I | R |
| QA vs pack | A/R | C | C |
| Publish | A | I | I |
Ambiguous ownership is how packs get ignored after kickoff.
Kickoff agenda (15 minutes)
- Read intent one-liner aloud (30s)
- Walk beat table jobs (5m)
- Show stills at phone size (4m)
- Highlight Do/Don't and claims (3m)
- Confirm QA gate + revision rounds (2m)
If kickoff runs an hour of taste debate, the pack was incomplete or unused.
After first delivery
Require a short vendor note: which beat IDs changed from pack v1, and why. Silent "improvements" are a common drift source—especially AI vendors regenerating whole clips when only captions needed a bump.
Red flags in vendor communication
- Asking for "full creative freedom" before reading the beat table
- Returning renders before board approval
- Changing host face for "better lighting"
- Expanding CTA into a hard pitch not in the pack
- Requesting a mega-prompt instead of beat-level notes
Answer red flags by pointing back to the pack section—not by rewriting the strategy in Slack.
Soft close
Treat the pack as the approval artifact before any vendor renders clips. Storyboard-first tools like Scro help you export that pack so freelancers grade against the same board you approved.