20cuts

Resources · By use case · 8 min read

Product Hunt launch videos: what actually works in the gallery

A good Product Hunt launch video opens on the product already running, works with the sound off, and puts its strongest moment inside the first ten seconds. Sixty to ninety seconds is enough.

Product Hunt is a daily leaderboard of new products, and each launch page has a media gallery where your video plays in the first slot. Visitors skim the day's list, open a few launches in tabs, and give each one a few seconds. Almost everyone watches muted, and many scrub through the video instead of playing it. So the video has to work visually, out of order, and fast.

Suppose the launch is Petal, an invoicing product: its 75-second video shows an invoice created, sent, and paid while an overdue list shrinks. Petal carries the examples from here on.

Four rules for the cut

Muted autoplay, scrubbing, and the few-second attention window set four rules:

Assume the sound is off

The gallery autoplays muted. So do the social feeds the video gets clipped into, and plenty of viewers sit somewhere they would not turn the sound on anyway. The picture has to carry the whole claim, and narration becomes a bonus channel for whoever opts in. Two practices follow:

Open on the machine running

A launch-day viewer clicked your tagline, so they already know the problem. A problem-first open, the right choice for a homepage explainer, re-sells someone who is already through the door. Open on the product mid-demonstration instead. Petal's tagline promises invoicing without the chasing, so Petal's video opens on an invoice getting paid without anyone chasing it, starting in second one.

The machine-running open: the whole product is frame one and a run is already underway. On launch day this is the open that fits, because the viewer arrives already knowing why they clicked.
The problem-first open, for contrast: it builds the pain before naming the product. Right for a homepage visitor who needs a reason to care, wasted on a gallery viewer who clicked your tagline seconds ago.

The whole cut compresses the hooks guide into four beats:

The launch cut, second by second
  1. 1Seconds 0–3. the product on screen, mid-demonstration; something moves in second one
  2. 2Seconds 3–10. one run, end to end, with the payoff staged large
  3. 3Second 10 onward. the mechanism, for viewers who stayed: one idea per scene, 8–12 seconds each
  4. 4The closer. what the viewer can do now; on launch day, usually one line: this is live, go try it
Sixty to ninety seconds total. The payoff sits where the audience actually is: the first ten seconds.

Staged large means large: a payoff smaller than about 1/40 of the frame did not happen, and the gallery thumbnail shrinks everything further.

Narration comes after the visuals are locked, and it stays light. The first ten seconds hold about one spoken sentence, so spend it on the claim rather than a welcome. Trailer voice fails here too: one draft in our corpus opened with a suspense line over epic music, and the review note called it marketing rather than explaining.

Launches that got it right

The clearest case is Screen Studio, which took #1 Product of the Day in February 2025 with 1,861 upvotes. The launch video is itself a Screen Studio export: the smoothed cursor, the automatic zoom on click, and the motion blur the product sells are the footage you are watching, so the payoff cannot land late. When your product makes videos, let it make the launch video.

Vibrantsnap, an AI screen recorder that also took #1 Product of the Day, opens its 72-second video on a finished demo video, the outcome, before showing the three clicks that produced it. The makers published their own breakdown, and their stated rule was to show the outcome first and the ease second. They report 40% of their launch traffic converting to trials.

Devin, the Cognition coding agent, launched in March 2024 on X rather than Product Hunt, and the demo clip that passed 30 million views obeys the same physics: a user types a prompt in the opening seconds, and everything after is Devin's own terminal, editor, and browser doing the job. It leans on a voiceover, so its muted survival depends on the on-screen activity carrying the story, which the terminal and editor changes mostly do.

Saidar, at indie scale, was singled out in a Product Hunt community roundup for the same pattern: the typed prompt is the storyline, and the product's response plays out from it. A prompt reads without audio and tells the viewer exactly what was asked.

Suno's v4.5 video is the one legitimate exception: the product generates music, so the video is built around sound. That exception is available only to products whose output is sound, and even then the visuals have to telegraph what is playing.

Famous videos that break the rules

Some famous launch videos break every rule here, and each one is instructive, because each was built for a different job than the gallery.

Cluely's April 2025 launch film is a scripted restaurant scene: a founder bluffs through a date with a hidden AI assistant, the date walks out, and the product's interface barely appears. As distribution it worked enormously, passing 13 million views and making the company a story. In a gallery it would fail everything: it opens on actors, it dies muted because the dialogue carries it, and a product payoff never comes. The viewer finishes knowing the controversy and not the interface.

Rewind's November 2022 launch won that year's Golden Kitty with a video of the founder talking directly to camera, the search demo arriving after the spoken setup. It worked because the founder's delivery and existing audience could carry a sound-on, full-attention viewing, and that is distribution most launches do not have. The rules here exist for products that have to win the muted, scrubbing viewer instead.

Humane's Ai Pin introduction is the full inversion: roughly ten minutes of two founders presenting to camera, with the first real payoff, the palm-projected display, arriving minutes in. Length, payoff placement, sound-dependence, and the open all point the wrong way, at maximum budget, and the tech press said so at the time.

The same failures show up in the gallery every week without the famous names: twelve seconds of logo animation before anything happens, or motion graphics that never show the interface. Product Hunt commenters are blunt about what they want, and it matches the rules: the interface, doing something real, quickly.

Three more gallery failures

Each maps to a rejection in our own review records; Product Hunt just punishes it faster.

FAQ

How long should a Product Hunt launch video be? Sixty to ninety seconds, with the strongest beat in the first ten. The length guide has the scene-count math; the launch slot adds only the requirement that the scenes are sorted payoff-first.

Does the video need a voiceover if everyone watches muted? Yes. A meaningful minority turns the sound on, and narration carries the why that the picture cannot show. The picture must never depend on it: write narration to the finished visuals, ship captions, and pass the mute test first.

Should the founder talk to camera? As a short frame it can work: twenty-five seconds or so, name the subject plainly, then hand off to the product running. Rewind proves a founder can carry a whole launch, and also proves what that takes: delivery and an existing audience most launches do not have.

Can we reuse our homepage explainer as the Product Hunt video? If it opens cold on the machine running, yes. If it opens with problem setup or brand build-up, cut a launch version from the same scenes with the payoff moved to the front. That is a re-sequence, not a new production.

You can skip the brief entirely: send your product's URL and pick from twenty short videos of it.

See the answer for your product instead of the average:

Get my 20 free videos