February 14, 2017 at 12:27AM


Converting 125fps to NTSC/30fps (FS7)

So I did the unthinkable - hit S&Q on the A cam just before a really important and one-off interview, cranking my FS7 up to 125fps.
Can anyone suggest is the best way to convert this back to useable 30fps / 24fps footage for broadcast? Ideally it needs to actually look good, not just useable.
(I was taking sound (two radio Mics) into the B cam so provided I can time sync the two after conversion I should survive on that front.)


Usable is the best you're going to get, sorry. The problem isn't the look so much as the sound.
You can slow down the video slightly to 120fps to get 30fps or 24fps. It's essentially the same as converting PAL to NTSC frame rates. The real tricky part is most video editors don't handle audio well. You may need to process it separately in an audio-specific editor to get optimal results. What you would do is slow the audio down to 46.0799KHz and resample it back to 48K 24-bit. The pitch will be lower and if you don't use a high quality sample rate converter, it will take on a shrill sound. You could instead try to use a time-stretch algorithm to change the length by 1.04167x without affecting pitch but if it isn't an extremely high quality one, you will hear a gritty or echoey sound.

February 14, 2017 at 7:18AM, Edited February 14, 7:19AM


I don't get what Stephen is talking about with the sound at all. Sound is sound. If you have a properly recorded audio file at 48K, praise God and get to work on retiming the video. Do do that, I agree that doing a conversion to 120 fps is the right first step, and then converting to 30fps (4:1 slowdown) or 24fps (5:1 slowdown) is trivial. Mate the audio and the video and you're done.

February 14, 2017 at 9:35AM


The camera would record no audio at all when shooting at 125 fps...

Page 43 from the US Sony FS7 Manual

"Audio recording is not supported in Slow & Quick Motion mode."

February 14, 2017 at 9:19PM

Guy McLoughlin
Video Producer

Guy and Stephen, audio was recorded on a second camera at 48k. That's the audio that will be used.

Michael Tiemann

February 14, 2017 at 9:29PM

Mr. Tiemann, if he slows the A-cam down to 120fps, it has been slowed by 4.17% so it will no longer be in sync with the B-cam, which contains the audio. The audio must also be slowed by the same amount to get it back in sync. The video from the B-cam can't be slowed because it's already at the target frame rate, so it's essentially useless except maybe for a very quick cut-away shot once in a while.

The only other option is to use a vector-based frame interpolation plugin to multiply the frame rate 6x to get 750fps, then divide down to 30fps. That would avoid slowing down the video and losing sync but as you probably know, would be unreasonably processor-intensive and has the possibility of errors. Forget about 24fps because the least common multiple with 125 is 3000!

February 15, 2017 at 7:40AM, Edited February 15, 7:47AM


Stephen, I understand the disconnect between our answers. Some frame-rate conversions involve simply playing at a different speed (i.e., you have 1250 frames shot at 125 fps, a 10 second clip, and you play them at 120 fps, which gives you a 10.417 second clip) while other conversions actually interpolate the answer (i.e., for every 25 frames of source, you play 24 of the frames, each one playing for 1/120th of a second rather than 1/125th of a second). I was proposing a two-stage interpolation (125 -> 120, then 120 -> 24 or 30), not a time-stretch followed by a decimation.

February 15, 2017 at 8:47AM


Ah, gottcha'. I have done similar but it can look jerky since some frames are displayed longer than others. Editors that support frame blending (or resampling as Vegas calls it) will allow a variable cross-fade between adjacent frames to alleviate that issues but it takes longer and can look blurry if the motion is very fast. At 120fps, I suppose the issue may be negligible.

Inexplicably Banned

February 15, 2017 at 1:23PM

Your Comment