Videly by the Numbers: What 222 AI Product Videos Tell Us

Between September 29 and November 3, 2026, 204 people made 222 product videos with Videly. The median video took 9.4 minutes to generate, 83% were made in 16:9, 89% used AI voice-over, and people asked for 45 rounds of edits in total. This post shares those numbers, how we measured them, and what they suggest about making product videos.

Most statistics about video marketing come from surveys of what marketers say they do. These numbers are different: they are what people actually did inside an AI video tool, measured by our product analytics. The sample is small and early, so read them as a snapshot, not an industry benchmark.

The numbers at a glance

How long does an AI product video take to generate?

The median Videly video took 9.4 minutes from clicking generate to a finished, rendered MP4, and 90% finished within 13.4 minutes. That time covers planning the storyboard, writing and timing the scenes, voice-over and music, and the final render. It does not include the time someone spends writing their prompt or collecting screenshots beforehand.

For comparison, a motion-designed product video from an agency or freelancer is usually measured in weeks, mostly spent on briefs and revision rounds. The practical difference is not only speed: when a first version takes ten minutes, it becomes reasonable to make one for every launch and feature instead of one per year. See how to make a product demo video for the full process.

Which aspect ratio do SaaS teams choose?

83% of videos were made in 16:9, 16% in 9:16 and only 1% in 1:1. Landscape still dominates for product videos because the first place they go is usually a landing page, a Product Hunt gallery, YouTube or a sales email, all of which are landscape-first.

Vertical is the clear second choice and is used for Reels, TikTok, Shorts and LinkedIn mobile. Square has almost disappeared. If you only make one version, make it 16:9 for your site; if you make two, add 9:16 rather than 1:1.

How many product videos use voice-over?

89% of videos were generated with AI voice-over, and every video had music. Voice-over helps a product video explain the why, not just show the screens. Because most social video is watched without sound, captions still matter: Videly adds them to every video so the story works on mute as well.

How often do people edit an AI-generated video?

Across 222 videos, people asked for 45 rounds of edits in plain language (for example "make the logo bigger" or "change the second scene's headline"). Most videos were accepted as generated; a smaller group of users iterated heavily, with one video going through 11 revisions. That pattern matches how teams work with designers: most assets are fine after a round of feedback, and the most important ones, such as a launch video, get polished many times.

How long are the videos?

Almost all videos in this sample were 15 seconds long. We do not read much into that: the Free plan caps videos at 15 seconds, and most people in the first five weeks were on the Free plan. Paid plans allow 30 to 60 seconds. For guidance on length by channel, see the landing page video guide.

What this means for your next product video

• Start with 16:9 for your website and launch pages, then cut a 9:16 version for social. Skip square unless a specific channel needs it.
• Use voice-over and captions together. Voice carries the story for people with sound on; captions carry it for everyone else.
• Plan for one round of feedback. Most videos need zero or one round of edits; save heavy iteration for launch videos that will be seen the most.
• Make more, smaller videos. At around ten minutes each, a video per feature release is realistic. Our feature announcement videos page shows the format.

Methodology

Data comes from Videly's product analytics (PostHog) for videos completed between September 29 and November 3, 2026. Team, internal and test accounts were excluded. Generation time is measured server-side from the start of a video job to its completed render. Format, voice-over and music settings come from the settings chosen when each video was started (212 first-version generations had complete settings). Edit rounds are completed plain-language edit requests on existing videos. We will update this post as the sample grows.