here seems to be some confusion on which mode will give you the highest quality video when recording 120fps video on the a6700. Movie mode can shoot at a maximum of 280Mbps, while the S&Q (Slow & Quick) mode can only shoot a maximum of 100Mbps. This has led some people to suggest using movie mode. The other side of it is people saying that if you do the math you get more data per frame in S&Q Mode, almost twice as much in fact, so you should use that. While the math made sense I decided to do an experiment myself to confirm.

Test Footage

I filmed two videos back to back shot with the same static content. One was shot in Movie Mode at 120fps, the other in S&Q (Slow & Quick) mode at 120fps with a base rate of 24fps. I typically use a 24fps base for my timelines when editing so this experiment is based around that. Also note that the 24fps isn’t 24fps, but 24000/1001 or 23.98fps, likewise 120fps is 120000/1001 or 119.88fps. I used the XAVC S 4K file format in both modes as this is the highest quality you can shoot with 120fps. I then set bit rate, color sampling and bit depth at 280M 4:2:2 10bit for Movie Mode and 100M 4:2:2 10bit for S&Q Mode.

I recorded for about 7 seconds in real time for each clip, then used ffmpeg to take the middle for evaluation. The files are in different frame rates so I’d need to use different second lengths in order to get close to the same total number of frames. The MovieMode file has a frame rate of 120fps so I can take three seconds with this command:

ffmpeg -i MovieMode.mp4 -ss 2 -t 3 -c copy MovieMode-3sec.mp4

The S&Q file has a frame rate of 24fps so I would need to capture five times as many seconds to get the same total number of frames 3 * 5 = 15. I can do that with this command:

ffmpeg -i SandQMode.MP4 -ss 9.265 -t 15 -c copy SandQMode-15sec.mp4

I can use ffprobe to get the average bit rate for each clip in bits per second, then pair it with awk and simple math to convert it to Mbps. Note that this is the video bit rate and excludes the audio. We do this because S&Q mode doesn’t record audio:

ffprobe -v quiet -show_entries format=bit_rate -of default=noprint_wrappers=1:nokey=1 video-file.mp4 | awk '{printf "%.2f Mbps\n", $1/1000000}'

We can also use ffprobe to grab the frame rate for the video file and use awk to format it:

ffprobe -v quiet -select_streams v:0 -show_entries stream=r_frame_rate -of default=noprint_wrappers=1:nokey=1 video-file.mp4| awk -F'/' '{if($2) printf "%.2ffps\n", $1/$2; else print $1"fps"}'

Finally I added a variable and echo to give me a cleaner output. I can now run this mega command against each trimmed file to get the information I care about:

f=MovieMode-3sec.mp4 && echo "$f" && ffprobe -v quiet -select_streams v:0 -show_entries stream=r_frame_rate -of default=noprint_wrappers=1:nokey=1 "$f" | awk -F'/' '{if($2) printf "%.2ffps\n", $1/$2; else print $1"fps"}' && ffprobe -v quiet -select_streams v:0 -show_entries stream=bit_rate -of default=noprint_wrappers=1:nokey=1 "$f" | awk '{printf "%.2f Mbps\n", $1/1000000}'

Output for file shot in Movie Mode:

MovieMode-3sec.mp4
119.88fps
287.08 Mbps

Output for file shot in S&Q Mode:

SandQMode-15sec.mp4
23.98fps
97.19 Mbps

Cool! The bit rates are different and align closely with what I set in the camera. The Movie Mode file is 195% more data per second than the S&Q file. That settles it, Movie Mode is better.

Wrong.

Slow it down

I could do some simple math to prove it, but I’ll keep manipulating the files to see the real world values. I’ll take the Movie Mode file and slow it down to 24fps. It’s important that none of the frames are re-encoded to keep the quality the same. Luckily there is a pretty easy way to do that with ffmpeg and -itsscale:

ffmpeg -itsscale 5.0 -i MovieMode-3sec.mp4 -vcodec copy MovieMode-15sec.mp4

This will multiply the PTSs (Presentation Time Stamp) so that they stay on the screen five times longer, or 23.98fps, but it will just copy the existing codec without re-encoding. I can then run the function from earlier against the MovieMode-15sec.mp4 file to get the bit rate and frame rate:

MovieMode-15sec.mp4
23.98fps
57.57 Mbps

Again for comparison the S&Q Mode file:

SandQMode-15sec.mp4
23.98fps
97.19 Mbps

Notice the difference now? The frame rates are the same, but the bit rates are not. In fact S&Q is about 69% more data per second than Movie Mode is. We can do some simple math to make some sense of this. S&Q mode is recording 120fps, but that is then being transformed in camera into 24fps at 100Mbps. The Movie Mode is recording 120fps at 280Mbps. If the intention is to slow this Movie Mode footage down to 24fps the new bit rate can be found by taking dividing the new frame rate by the older frame rate and multiplying it by the old bit rate:

24fps/120fps * 280 Mbps = 56 Mbps

That’s pretty close to the 57.57 Mbps that I got on the real world files.

Conclusion

If you want the highest quality 120fps video that you are going to slow down to 24fps AND you don’t need audio then use S&Q Mode. If you need audio OR you want the highest quality for playing back at greater than 30fps you should use Movie Mode. Is there a big subjective difference in the image quality of one vs the other? I’m not sure as I have tested both enough side by side to see. For my use cases I’ll likely just shoot in S&Q as I rarely need 5x slow-motion, will probably use it for B-Roll, won’t need the audio and can handle larger file size.