Video frame rate: 24, 30 or 60 fps and what conversion does to your clip

Which frame rate should you export at?

In most cases, export at the rate the footage was recorded at. A plain conversion, such as the ffmpeg fps filter we tested, can only repeat or drop frames that already exist, so it adds no detail; it only changes which of your frames are on screen and for how long. Frame blending and optical flow do create new in-between frames, which can look smoother but can also smear or warp moving edges, and we did not measure them.

There are three common exceptions. On a timeline that mixes rates, you convert the odd clips to the timeline rate. Slow motion is recorded at 60, 120 or 240 fps and played back at a lower rate on purpose. And a client, broadcaster or festival may name a delivery rate, which then wins.

Choose the recording rate before you shoot, from the motion you will film. This table is our suggestion, not a platform rule, and the only part we measured is what conversion does (next section).

YouTube publishes 8 Mbps as the recommended 1080p SDR video bitrate for 24, 25 and 30 fps and 12 Mbps for 48, 50 and 60 fps, read on YouTube Help on October 27, 2026. Our arithmetic (script in the test log): 12 / 8 = 1.50, so 50% more. These are recommendations, not requirements.

What does converting the frame rate do to your frames?

To see it, we took 10 second excerpts, starting at 60 seconds, of three real versions of Big Buck Bunny: a 24 fps file, a 30 fps render and a 60 fps render, all with distinct frames in our excerpt. We converted each to 24, 30 and 60 fps with ffmpeg's fps filter, which repeats or drops frames, then hashed every decoded frame and matched each output frame back to the source frame it came from.

The control was the same-rate conversion (24 to 24, 30 to 30, 60 to 60). Every one of them kept all source frames, repeated nothing and dropped nothing, as it should. Before running, we expected 30 to 60 to repeat every frame once (50%) and 24 to 60 to alternate holds of 3 and 2 (60%); the log matched.

Our measurement, October 27, 2026, ffmpeg 6.1.1, Big Buck Bunny excerpts (Blender Foundation, CC BY 3.0). A hold is how many output frames show one source frame in a row: 0 means that source frame never appears. We compared frames as 160x90 grayscale hashes of the decoded pictures, with no encoder in between, so these numbers describe the frame selection and not what a platform's own encoder does afterwards.

Two patterns matter. Going up, 30 to 60 gives an even hold of 2 for every frame, while 24 to 60 alternates 3 and 2, so some frames stay on screen 50% longer than their neighbors (3 against 2, computed). Going down, the frames are simply gone: 60 to 30 never shows 300 of the 600 frames that were recorded.

We did not run a viewer test, so we make no claim about how visible the uneven hold is. It depends on the motion in the shot. In mostly static footage you will not see it; in a steady pan you may.

Does real camera footage behave the same way?

Yes. Animation is the easy case for a frame test, so we repeated it on two live-action clips: a 30 fps phone clip of a bird on a hand (CC0) and a 23.976 fps clip from a Sony FS700 camera (CC BY 3.0), 10 seconds each, every source frame distinct. The same-rate control again kept every frame and repeated none.

Our measurement, October 27, 2026, same script method (test-live.py), ffmpeg 6.1.1. The phone clip gave exactly the animated film's numbers. The camera clip shows what "24 fps" usually means in practice: 23.976 fps, so converting it to 30 repeats 61 of 301 output frames, one more than a true 24 fps file.

Converting 23.976 to 24 fps changed nothing in our 10 second excerpt. The two rates differ by 0.1%, so the fps filter repeats one frame about every 1,001 frames, roughly every 41.7 seconds (our arithmetic: 23.976 fps is exactly 24,000 / 1,001, so 24 / (24 - 24,000 / 1,001) = 1,001 frames, which at 24 frames a second is 41.7 seconds).

How do you check and set the frame rate yourself?

Read the video stream with ffprobe. In our check, a 10 second excerpt of the 24 fps source printed r_frame_rate=24/1 and nb_frames=240, and the same file after conversion printed 30/1 and 300 frames, the numbers we expected before running it.

Convert only when you must, for example to match a mixed timeline. The fps filter is the plain repeat or drop conversion that we measured; editors may offer frame blending or optical flow, which we did not measure.

If r_frame_rate and avg_frame_rate differ, the file may have a variable frame rate, which many phones produce. It is a sign, not proof: some constant-rate files also show two different values, and a variable file can show equal ones, so check the frame timestamps (ffprobe -show_entries frame=pts_time) or a tool such as MediaInfo before you convert. Our test files were constant. Convert such a file to a constant rate in your editor before you cut it, and check the result with the command above. We did not test a variable frame rate file for this page.

ffprobe -v error -select_streams v:0 -show_entries stream=r_frame_rate,avg_frame_rate,nb_frames -of default=nw=1 input.mp4

ffmpeg -i input.mp4 -vf fps=30 -c:v libx264 -crf 20 -pix_fmt yuv420p output.mp4

What do YouTube, Instagram and TikTok publish about frame rate?

YouTube Help (read October 27, 2026) says: "Content should be encoded and uploaded in the same frame rate it was recorded." It lists common rates of 24, 25, 30, 48, 50 and 60 frames per second, says other rates are also acceptable, and asks for interlaced content to be deinterlaced first, for example 1080i60 to 1080p30.

The Instagram Graph API reference for Reels (read October 27, 2026) lists "Frame rate: 23-60 FPS", MOV or MP4, HEVC or H264, progressive scan, and a 3 second to 15 minute duration. That is the specification for uploads through the API, not a description of what the Instagram app does to a file.

The TikTok Content Posting API guide (read October 27, 2026) lists "Minimum of 23 FPS" and "Maximum of 60 FPS" for video, in MP4, WebM or MOV. Like the Instagram figure, it is the API specification and not a description of what the TikTok app does to a file.

Limits and what we did not test

• We tested one animated film in three versions and two live-action clips: a 30 fps phone clip and a 23.976 fps camera clip. We did not test noisy low-light footage or a variable frame rate phone recording.
• The 30 and 60 fps files are separate downloads of the same film, not one file converted by us. The control for each source was a same-rate conversion, which gave 0 repeats and 0 drops.
• We tested ffmpeg's fps filter only. Frame blending, optical flow and an editor's own conversion were not measured.
• We did not upload anything to YouTube, Instagram or TikTok, so we do not know what each platform's encoder does to a 24 fps or 60 fps file. The platform statements above are quoted from their help pages.
• We did not test 25 or 50 fps, drop frame timecode (29.97 and 59.94 fps) or 120 fps footage, and we did not run a viewer test of how visible the uneven hold is.
• The free Inspect tool on this site measures the frame rate by playing the first frames of the file in your browser; a file faster than your screen's refresh rate reads low there, so check 120 fps footage with ffprobe as shown above. No Koldflux free tool changes the frame rate.