Worked scripts as production docs (shot · on-screen text · spoken · sound · timing), with a
1-second hook, a loop, and TikTok-search keywords — plus a "polished ad" version for contrast. Voice
is the blunt indie-founder voice from earlier skills (short, dry, no exclamation marks, concrete).
Your output uses the loaded voice.md.
Example 1 — Storytime / talking head (filmed) · ~30s · goal: reach + saves
Concept: we deleted our most-requested feature; the boring button won. Sound: original audio (your voice). Format: talking head, handheld.
| Time | Shot | On-screen text | Spoken |
|---|---|---|---|
| 0:00–0:02 | Close, mid-sentence, looking at camera | "We deleted our #1 requested feature." | "We spent six months on the feature everyone asked for. Then we deleted it." |
| 0:02–0:10 | Same, quick cut | "nobody used it" | "Every sales call, people begged for it. We shipped it. Nobody touched it." |
| 0:10–0:20 | Slight push-in | "the boring button won" | "What they actually used was a tiny export button we kept ignoring. Built it in two days. Most-used thing we have." |
| 0:20–0:28 | Cut, direct | "watch what they DO" | "People tell you what sounds good and do something else. So now we watch behavior, not surveys." |
| 0:28–0:30 | Back to opener framing | "so we deleted it." | "Which is how we ended up deleting our most-requested feature." ← loops to 0:00 |
Loop: the last line flows back into the first frame — it replays seamlessly.
TikTok SEO: "feature", "SaaS product", "build" spoken + on-screen; caption: "How we decide what
to build for our SaaS — behavior over requests." Keyword + niche hashtags (#buildinpublic #saas
#productmanagement), not generic ones.
Caption + hashtags → caption-writer. Schedule → scheduling-and-queue.
Example 2 — Faceless listicle (AI b-roll + voiceover) · ~20s · goal: reach + saves
Concept: 3 tools that replaced a $200/mo stack. Format: faceless, voiceover over b-roll. Sound: trending upbeat sound, cuts synced to the beat.
| Time | Shot (Veo b-roll) | On-screen text | Voiceover |
|---|---|---|---|
| 0:00–0:02 | Fast push-in on a cluttered dashboard | "3 tools that killed my $200/mo stack" | "Three tools that replaced two hundred dollars a month." |
| 0:02–0:08 | Tool 1 in use | "1. ___ (free)" | "One — handles all my scheduling. Free tier covers it." |
| 0:08–0:13 | Tool 2 | "2. ___" | "Two — does what I paid fifty bucks for." |
| 0:13–0:18 | Tool 3 | "3. ___ (the sleeper)" | "Three — the one nobody talks about. Wait for this one." ← open loop paid off |
| 0:18–0:20 | Back to first frame | "save this 👇" | "Save it before you renew anything." |
Loop: "save this" + the fast pace invites a rewatch to catch all three.
Veo note: b-roll only; per veo-prompt-pack.md, this is a fair AI use (faceless, cutaways) —
but a founder review to camera might out-trust it. AI-disclosure auto on TikTok.
Contrast — the "polished ad" version (do NOT do this)
- 0:00 a 2-second animated logo sting.
- A voiceover in brand-announcer tone: "At [Company], we believe…".
- Glossy, color-graded, no face, tagline card at the end.
→ Reads as an ad in the first second; scrolled. On TikTok, the handheld founder-to-camera version beats it every time. If a user asks for the ad version, redirect to native.
What the good examples share
- A 1-second hook on all three channels (verbal + on-screen text + visual).
- Fast pace, no intro tax, no dead air.
- A loop / rewatch trigger built into the ending.
- A sound choice named (original vs trending) and, when relevant, beat-synced.
- TikTok-search keywords in spoken + on-screen + caption.
- Native, not ad-like; on-voice; AI-disclosure where generated.
