A feature release video is a short explainer, usually 45 to 90 seconds, that ships next to a changelog entry and teaches one new capability to people who already use the product. The economics only work as a series: the first video pays for a standing visual system, and every video after it mostly reuses that system.
Most teams run this backwards. They spend the budget on one big launch video, then ship the next eight months of features as changelog text nobody reads. The launch video was the expensive kind. Release videos are the cheap kind — if the cadence is designed for from video one.
Take an analytics product shipping a new funnel view. Its release video shows events streaming in, a funnel chart filling stage by stage, and the drop-off between two stages getting clicked open. That feature carries the examples below.
A release video is not a launch video
| Launch video | Release video | |
|---|---|---|
| Audience | Strangers deciding whether to care | Users who already have a mental model |
| Job | Excitement: the machine running end to end | Comprehension: one new capability slotted into an existing model |
| Altitude | The whole story | One feature, at "what it does and when you'd reach for it" |
| Length | 60–90 seconds | 45–90 seconds, short end usually right |
| Frequency | Once or twice a year | Every meaningful release |
The audience row drives everything else. A release viewer already knows how charts work in your product, so the video says: a funnel is a chart whose stages know about each other. It never re-explains the product, and it stays calm — the viewer already cares, and a hype edit on a settings improvement reads as insecurity. The once-a-year excitement piece is the launch video.
Refuse most changelog lines a video
A cadence stays sustainable when most changelog entries get no video. We keep a standing "cut, never animated" list for exactly this: UI procedures, settings walkthroughs, permission tables, small fixes, and reference facts, which only ever earn a screen recording or a docs paragraph. Route each entry with three questions:
- Does it change the user's causal model? A new object, a new behavior, a thing that used to be impossible — that earns a video, because animation exists to stage what a screenshot cannot show. The funnel view qualifies: new object, new behavior.
- Is it a procedure? "Here is where the new button lives" is a 15-second screen recording, captured in an afternoon. Animate the concept, record the live product.
- Is it a fix or a tweak? Changelog text. A video for it trains your audience that your videos are skippable.
Most products land at a video for one release in three or four. The videos stay dense, and the audience learns that a video means something actually changed.
Give each release one claim
Each video argues a single declarative sentence. For the funnel view: you can see exactly where users leave, and click the place they leave from. The lazy default — "this video is about the funnel feature" — produces a feature tour, and a feature tour has no strongest moment. Test the sentence the same way a launch claim gets tested: falsifiable, served by every scene, framed as a capability rather than a count. "The five new operations" dates the moment you ship a sixth.
Ground every value on screen, and re-derive all of them at every build. The product just changed, which means any number borrowed from a previous video is the most likely thing on screen to be wrong. The grounding method is one email to whoever operates the live build, sent before the script locks.
The second video costs half
A finished explainer sits on a pile of decisions: a visual language for what a chart is, timing constants, a camera grammar, a voice, a music bed, a title treatment. A one-off pays for all of it and throws it away. A series pays once and freezes it:
- Motion timing is decided once. Reveal durations, camera moves, and cause-and-effect beats live in a tuned table every video imports, mined from graded-good renders (the timing constants).
- Voice settings are tested once, then pinned. Approved audio takes get kept and reused rather than re-synthesized, because regenerating "the same" line produces a different take.
- Surfaces promote on second use. Anything two videos both need — the chart frame, the title card — becomes a shared component the third video gets for free. Promotion waits for the second use: a library built before any reuse is speculative, and one built after five copies has drifted.
- Words stay the cheap layer. Narration is written to the finished picture and synthesized from text, so a wording fix on a shipped video is a 60-second loop: edit the line, re-synthesize one scene, re-render.
By the third or fourth video, production is a new worked example, a new claim, and the same rig. Per-video pricing from a vendor with no standing system understates your real cost, because you are re-buying the rig every time, and it comes back slightly different every time.
Five rules keep the series consistent
- Each video owns one idea and defers its neighbors. This video owns what a funnel is; its neighbors own segments and alerts. When two videos half-teach the same concept, both fail.
- No "welcome back." Release videos get watched months apart, out of order, from a docs page or a tweet. Each one self-contains its prerequisites in one line and gets to the point.
- One home for every fact. Brand strings, product claims, and pricing get one source of truth that every script points at. The same fact written independently into two videos is tomorrow's contradiction.
- "Redo" is a first-class state. Keep a plan of record marking each video exists, redo, or new. A stale release video demos a product that no longer exists, which is worse than no video.
- Founder intros stay short and framing-only. 25 to 30 seconds, a plain noun open, two or three concrete use cases, a hand-off line — and never a line the video itself repeats.
Vary the staging even though the rig stays fixed. One technically clean build in our records repeated the same good beat shape five times and drew a one-word rejection: "boring" (the rhythm story).
FAQ
How long should a feature release video be? 45 to 90 seconds, short end when in doubt (the length math). A script that wants three minutes is either two videos or a launch video.
How fast after the release should it ship? Same week, once the code is merged and a real run exists to ground against. With the standing system in place, a release video is a worked example and a script away.
Can we reuse scenes from the previous video? Reuse the system: components, timing, camera grammar, voice. Never reuse rendered scenes with values swapped — the release you are announcing is precisely what made the old values stale.
Do release videos work without an existing audience? They compound with one, but their first home does not need one: the changelog entry, the feature's docs page, the announcement post. The same video does quiet duty in onboarding once the feature stops being new.
If you want this done for your product, start with the URL. We come back with twenty short candidates and you pick from the board.