From brief to ship — a drafting loop

Open /draft. Paste a plain-English brief (product, audience, CTA, vertical, format). You get 6–10 hook variants, a 3–7 cut shot list, one full sample script you could shoot today, and risk flags specific to the vertical.

Each hook is a distinct angle — pattern interrupt, curiosity gap, social proof, problem-agitate-solution — not permutations of one line. The sample script ties to the strongest hook in the list.

Run cost: ~15 seconds, charged against the 60/day team draft quota.

If you're hiring UGC creators, head to /draft/creator-brief instead of the generic /draft. Paste brand + product + tier + budget + count → N distinct paste-into-DM brief bodies, each tilted to a different angle so the resulting cuts aren't carbon copies.

Save successful brief inputs as templates for reuse next campaign — the picker bar at the top of the workbench loads any saved template into the form with one click.

Got 2–4 concepts from drafting (or from your team's past campaigns)? Drop them into /compare/concepts. Side-by-side scoring on hook + potential, strengths/weaknesses per concept, predicted winner with rationale, concrete next step (combine, A/B test, kill candidate X).

The Compare surface is faster than running full Deep Bench on each concept — useful for the "which one do we shoot first" decision.

Once you have a winner — either from compare or from a real-world campaign — open /draft/variants. Paste the winning script + its TokBench score, pick a transformation axis (hook-only, CTA-only, full rewrite, format-shift, tone-shift), and get up to 10 variants that preserve the winning structure while varying along the chosen axis only.

Each variant has a one-click button to run it back through the analyzer. This is the loop that turns one winning ad into a creative system: score → variants → score → variants of the new winner.

Sometimes you don't need full variants — you just need to rank 2–4 hook openings. /draft/hook-test takes opening lines (first 2 seconds), ranks them by predicted hook rate, and explains why each will perform as predicted. ~5 seconds.

Use it as a low-friction filter before committing to a full shoot. Run the winner through Deep Bench afterward.

Before any ad goes into production, run it through /draft/compliance. FTC + TikTok Shop + general claim flags with quotes, severity grades, and concrete fixes. Status banner: clean / warn / block. When score is below clean, you also get a revised script with all issues corrected — paste-in ready.

Skipping this step is how teams get ads rejected mid-campaign. Spending 5 seconds here is cheaper than a re-shoot.

When a UGC creator submits a script for your team to approve, drop it into /draft/voice-match. 0–100 score against your team's Brand Kit voice doc, dimension breakdown (tone, banned words, audience fit, CTA style), off-brand line quotes, and a revised script when score < 80.

Pass = ship. Review = revise. Fail = reject and brief again. Set a team standard at /account/team/brand-kit first.

Once the script is locked, generate cover-frame prompts at /draft/thumbnail. Three distinct visual angles per script, each compatible with DALL-E / Flux / Midjourney. One-click copy per prompt.

The cover frame is your real first impression before autoplay kicks in. Don't leave it to whatever the editor screenshots.

After a successful campaign, save the winning hook to your team's hook library at /account/team/hooks. Tag by pattern (curiosity gap, social proof, etc.) and vertical. Sortable by score.

Six campaigns in, the library becomes a searchable playbook of your team's greatest hits — far more useful than scrolling past Slack history every time someone asks "what was that hook we used last quarter?"

Brief → drafts → compare → pick → variants → score → compliance → voice-match (if UGC) → thumbnail → ship → save the winning hook → next brief.