October 6, 2026
24, 25, 30 or 60 fps: which frame rate to choose for an AI video
Why cinema runs at 24 frames per second, Europe at 25 and America at 30, which frame rate Seedance, Kling, Veo or Wan output (measured on our files), and which one to aim for on YouTube, TikTok, Reels, television or a display screen.
Frank HoubreFounder of Imaginode
24 frames per second: a 1926 compromise
Cinema has run at 24 frames per second since sound arrived: in 1926, Western Electric set the speed of sound film at 90 feet per minute, which is 24 frames, the average projection speed of the time.
Before sound, silent films ran at 16 to 22 frames per second, and projectionists sped them up as they pleased. Sound recorded on the film required a fixed speed. Western Electric engineers settled on 90 feet of 35 mm film per minute, which is 24 frames per second, because that was roughly the speed theatres were already projecting at. The Jazz Singer, in 1927, set that figure for a century.
That economic compromise became an aesthetic. At 24 frames per second, with the shutter open half the time, each frame carries a slight motion blur, and camera moves keep a small judder. That is what the eye associates with cinema. Raising the frame rate makes the picture sharper and smoother, but also closer to news footage or sport.
So it is no accident that almost every AI video model adopted 24 frames per second: they learned from films, ads and music videos, and they aim for that look. The rest of this article says when to keep that frame rate, and when to change it. The conversion itself is covered in convert a 24 fps video to 30 fps.
25 and 30: a legacy of the power grid
Television tied its frame rate to the mains frequency: 25 frames per second in 50 hertz countries, Europe, Africa, Australia, most of Asia; 29.97 in 60 hertz countries, the Americas, Japan, South Korea, Taiwan.
Early TV sets synchronised to the mains to avoid flicker. At 50 hertz, that meant 50 fields, so 25 frames per second: PAL and SECAM, in Europe, Africa, Australia, much of Asia and the Middle East. At 60 hertz, 30 frames, which became 29.97 with NTSC colour: North America, most of Latin America, Japan, South Korea, Taiwan, the Philippines. Brazil is the exception, PAL but at 29.97.
High definition kept the same split, and it still matters to anyone delivering to a channel. France Télévisions' technical annex of January 2023, for example, requires HD files at 25 frames per second, interlaced, at 50 hertz. A shot generated at 24 must therefore go to 25 before entering a programme broadcast in France.
Online, that split fades: a computer or phone screen usually refreshes at 60 hertz or more, everywhere in the world, and shows any frame rate. But habits remain: European creators often shoot at 25 or 50, American ones at 30 or 60, and an edit that mixes both has to choose.
AI videos almost all come out at 24
Measured on 6 October 2026 on the raw files of our account: Seedance 2.5, Seedance 2.0, Kling V3, Veo 3.1, Flux 3 Video, Grok Imagine and Hailuo output 24 frames per second; Wan 3.0 and Wan 3.0 Prime output 30.
We read back, with the ffprobe tool, the videos actually produced by each model in the catalogue, as they arrived from the provider. Seedance 2.5, Seedance 2.0 and its Mini version, Kling V3, Veo 3.1, Flux 3 Video, Grok Imagine Video and Hailuo all render 24 frames per second, whatever the resolution. Google says so in the Veo documentation: one supported frame rate, 24.
The exception is Wan 3.0, and its Prime version, which output 30 frames per second, at 480p as at 1080p. An edit that combines Wan shots and Seedance shots therefore mixes two frame rates, and editing software resolves that conflict silently, by duplicating or skipping frames.
None of these models lets you choose its frame rate: it is a fixed provider setting. To deliver at 25, 30 or 60, conversion happens after generation, shot by shot. In the canvas, that is the job of the Enhance node, which targets 24, 25, 30, 50 or 60 frames per second by creating the missing frames.
YouTube: keep the original frame rate
YouTube asks for video to be uploaded at the frame rate it was recorded in, and lists 24, 25, 30, 48, 50 and 60 frames per second among the common rates; an AI shot at 24 can go as is.
YouTube's encoding help page is clear: content should be encoded and uploaded at the frame rate it was recorded in. It names 24, 25, 30, 48, 50 and 60 frames per second as common rates, and accepts others. A video generated entirely at 24 therefore has no reason to be converted for YouTube.
The question comes up as soon as sources are mixed. A creator who films their talking head at 30 or 60 and cuts in AI shots at 24 gets, at export, AI shots converted by duplication: they stutter slightly on pans, next to perfectly smooth filmed footage. In that case, convert the AI shots to the shooting frame rate before editing.
For a gaming or screen tutorial channel shot at 60, the same logic applies: you do not drop the whole edit to 24 for a few AI shots, you bring the AI shots up to 60. The look stays smooth from start to finish.
TikTok, Reels and Shorts: 30 is the safe choice
The TikTok and Instagram APIs accept 23 to 60 frames per second, but Instagram's help page mentions a minimum of 30; delivering at 30 avoids any conversion by the platform, and 24 is still accepted.
TikTok's developer documentation states a minimum of 23 and a maximum of 60 frames per second. Meta gives the same range for Reels in its Instagram API documentation, but the consumer help page on Reels mentions a minimum of 30 frames per second. YouTube has no Shorts-specific guidance: the general rule applies.
None of these platforms officially says how it handles a 24 fps video. What we know is that they re-encode everything they receive. Delivering directly at 30 removes the uncertainty: the frame rate is within everyone's range, and no frame duplication is added on top of compression.
In practice, for a vertical video made of AI shots, two choices make sense. Keep 24 for the cinema look, especially on slow shots, portraits, products. Go to 30 as soon as there are pans, drones, scrolling text, or phone footage, which is almost always at 30. The picture size to aim for on each platform is in which format and resolution to generate.
Television, streaming, cinema
A European channel wants 25, a North American channel 29.97, Netflix the native frame rate of the production, and a cinema 24: for these deliveries, the frame rate is set by the specification, not by taste.
For television, the delivery spec rules. In France as in the rest of Europe, it is 25 frames per second, often interlaced. An AI shot at 24 gets there in two ways: sped up by 4.17%, the historical method for films on television, or converted by interpolation if the duration must stay exact.
Netflix requires video to be delivered at its native frame rate, that of the shoot and the edit, and warns that improper conversions, which leave artefacts, are rejected at quality control. For a film made of AI shots, the native frame rate is the models', 24 in most cases: edit at 24 and deliver at 24.
For a theatrical or festival screening, 24 is the norm. The platforms that accept AI films and their requirements are covered in where to publish your AI film.
The case of 60 frames per second
60 frames per second is for video games, sport, screen tutorials and display screens; on a fiction shot, it gives a very smooth video look that many viewers associate with TV motion smoothing.
60 makes everything sharper in motion: a ball, a car, scrolling text, an interface. It is the frame rate of video games and screen captures, and of many sports channels. For an in-store screen or a shop window looping a rotating product, it brings a softness that 24 does not have.
On a fiction shot, the effect is more debated. Motion blur disappears, the scene looks live, and that is exactly what filmmakers object to in TV motion smoothing. Converting an AI short film to 60 to make it smoother often makes it less cinematic.
60 also has a cost: twice as many frames to compute, so a doubled price at interpolation providers, and heavier files. Keep it for uses where smoothness is the goal. What interpolation gets right and wrong, frame by frame, is shown in AI frame interpolation.
The decision table
Film, festival, streaming platform: 24. European television: 25. Social media with motion: 30. Edit with filmed footage: the shooting frame rate. Games, sport, display screens: 60.
If the whole project is made of AI shots and ends up on YouTube, at a festival or on a platform: stay at 24, the models' frame rate, and convert nothing. If a Wan 3.0 shot at 30 slips in, bring it down to 24 rather than the other way round.
If the project mixes AI shots and filmed footage, the shooting frame rate decides: 25 or 50 in Europe, 30 or 60 elsewhere and on phones. Convert the AI shots to that frame rate, shot by shot, before editing. If the deliverable has a spec, a channel, an ad agency, a screen, the spec sets the frame rate.
In Imaginode, conversion happens in the Enhance node: 6 credits per five seconds at 720p up to 30 frames per second, 12 at 50 or 60, duration and sound kept. The ready-made workflow is the "Change a video's frame rate" template.