rollProduct Hunt launches are decided in the first hours, by people skimming. Your gallery's first asset does more work on that page than your tagline, your description, and your comment thread combined, because it's the only thing that plays while someone is still deciding whether to read anything.
Most launch videos get this backwards. They open with a logo animation.
The three seconds that matter
On the leaderboard, a hunter's attention arrives already spent. Whatever plays first has roughly three seconds to answer one question: what is this, and why would my life be different?
That means:
- No logo card. Your brand mark at the top is a three-second ad for a name nobody knows yet. Put it at the end, where it becomes the thing they remember.
- No slow fade in. Start on motion that's already happening.
- No "introducing." The word costs you a second and communicates nothing.
Open instead on the single most interesting thing your product does, mid-action. If your product finds cheap flights, open on a price dropping. If it monitors servers, open on the alert firing. The viewer will infer the category faster from one concrete moment than from any explanatory sentence.
Format: what survives the gallery
Product Hunt renders gallery media in a horizontal frame, and most people browse on desktop during launch day but share on mobile after. Practical consequences:
- Lead with a 16:9 cut for the gallery itself. A vertical video letterboxed into a horizontal slot wastes two-thirds of the frame.
- Keep a 9:16 version ready for the X and LinkedIn posts you'll write within the hour, and for anyone re-sharing to Stories.
- Assume it plays silently. Gallery autoplay is muted, so anything essential said out loud must also be legible on screen.
- Stay under 60 seconds for the gallery piece and under 30 for social. Watch-through matters more than completeness.
That "assume silent" rule is the one most launch videos violate. If your narration carries the value proposition and the on-screen text is just your logo, a muted viewer learns nothing.
Structure that works
A reliable shape for a launch video, roughly 30 to 50 seconds:
- Hook (0–3s). The interesting moment, mid-motion.
- Problem (3–8s). The frustration, in your user's words, not your architecture's.
- Reveal (8–20s). The real interface doing the real thing. Actual pixels from your actual product.
- Proof (20–35s). The number, the testimonial, the integration list: whatever is true and specific. One of these, not all three.
- Payoff (35–45s). Brand mark, domain, one instruction.
Two notes from watching a lot of these get cut. First, proof is where honesty pays: a real "312 teams" beats an invented "thousands," because specificity reads as true and round numbers read as marketing. Second, the reveal should show one workflow completing, not five features existing. A viewer who understands one thing completely will click. A viewer shown five things understands none.
The night-before checklist
Launches are usually assembled at midnight, so here's the version you can actually run under pressure:
- Video opens on motion, not a logo. ✅
- Product's real UI appears within the first eight seconds. ✅
- Every spoken claim also appears as on-screen text. ✅
- One number, and it's true. ✅
- Ends on domain + one instruction. ✅
- Exported horizontal for the gallery, vertical for social. ✅
- No licensed music. Original or generated audio only, because a copyright claim on launch day is a bad day. ✅
- Under 25MB so the upload doesn't stall at 3am. ✅
Where this gets automated
Everything above is a set of directing decisions, and directing decisions can be derived from the thing you're launching. That's what Rollsee does: it reads your product page, films the live site in a real browser, decomposes the interface into pieces it can animate, and storyboards a cut using only claims your own site makes. It scores the cut with generated music, so there's nothing to get flagged, and renders both the 16:9 gallery cut and the 9:16 social version from the same timeline.
If the pacing isn't right for your launch, you say so in plain words and it re-cuts around the note. Watching the first cut is free; you pay only when you want the watermark-free file. There's a feed of fresh cuts if you want to judge the output before you paste anything.
The launch-day version of all this fits in a sentence: open on the product doing something, prove one true thing, end with the domain. Everything else is decoration, and decoration is what you cut when the clock says 2am.