Suggestion for smaller file size with good quality

[Neubivljiv]

Similar questions have already been asked, but I could not find a solution.
In some of previous version (long time ago), I remember that I used all default export option. And it was perfect. Small file size, very good quality.
But after some updates, whatever I do, or I have good quality but a huge file, or vice versa.
I tried various options with codecs, rate control, Bitrate but I get either a huge file with good quality, or a smaller file with very poor quality.
For example, I got 10 min video and with good quality file size is almost 1GB. I remember that with earlier version I could export 10 min video (same resolution) who had ~120MB or less with good quality. And I dont even touch export options.
Now it seems impossible or I am missing something.

[Austin]

Without knowing specifics, here are some general suggestions:

1. Restart Shotcut to get back to the default export option.
2. Ensure that Export panel > “Use hardware encoder” is not checked unless you have a 7th-gen NVENC GPU like Nvidia RTX 20xx or higher.
3. Hit the Advanced button, go to the Codec tab, and lower the quality percentage until the quality/file size trade-off is acceptable to you.
4. If step #3 was not sufficient reduction, go to the Advanced > Other tab, change these lines to say vprofile=veryfast and vpre=veryfast, then repeat quality tests for step #3.

I’m guessing that hardware encoding could be causing the file size problem. Older GPUs frequently produce file sizes that are 3x larger than software encoders for the same quality level.

[kagsundaram]

I ran some tests on this recently.
Here is a summary; follow the link for the details.

@kagsundaram

My procedure then was to change each parameter one at a time, restoring it to original before changing the next parameter.

• Frame rate
• Bframes
• GOP size
• Quality
  Then a final test with all of these set to the same size-reducing values as the above tests.

@kagsundaram

changing the Bframes upwards from 3 to 8 had a less than 1/2% improvement in file size. So we can discard that idea.

Each of the other changes improved the file size; yet there was no noticable quality degradation of the Exported video.
Using all optimizations reduced a file size of 85.4 MB to 31.7 MB.

[Austin]

There are a few caveats:

• Changing frames per second will radically affect the illusion of motion (ie, cause stutter). This is generally not an option except for surveillance footage and PowerPoint slides. I’m sure you already know this. I’m just explaining why I don’t include it as a general recommendation in my replies.

• B-frames can cause a dramatic reduction in file size for videos involving camera movement or moving objects. Talking head footage, not so much, because so little is changing between frames. Diminishing returns happen at 8 B-frames. Hardware players often don’t support more than 3 B-frames.

• GOP is 5xFPS on the default setting, which is 150 for a 30fps project. The H.264 specification maximum is 250. The file size reduction from 150 to 250 is generally very small (diminishing returns). Aside from larger seek jump intervals when scrubbing playback, it doesn’t hurt to try up to 250 though.

• This leaves quality percentage (CRF) as the main driver for general recommendations. To go beyond this requires the user to really know what they’re doing to customize settings for their specific footage traits.

[Neubivljiv]

Hi, Austin.

Thanks for suggestion. I cut video to 2 min for testing and got some results. File size is ~42MB, good quality. My GPU is 1080 GTX and earlier I almost always used Use hardware encoder.

Simple solutions always seem to be the best.

Best regards.

[Austin]

Thanks for replying with the new file sizes. It helps to see the impact of different settings for reference by future users.

[kagsundaram]

@Austin

B-frames … Talking head footage, not so much, because so little is changing between frames.

Thanks.
This clearly explains my results. (It was talking-head footage.)

@Austin

GOP … file size reduction from 150 to 250 is generally very small

It must vary with the video image type, motion etc. I got a very significant size reduction by increasing GOP on talking-head footage.

[Austin]

You are correct, the type of footage affects the efficiency of GOP. Talking head can potentially benefit from longer GOP because changes are very minor between frames, so there’s no reason to burn disk space by writing an entire new I-frame of the same thing. However, “talking with hands” or showing products in front of the camera can blow those efficiencies apart, hence the falloff around 150 GOP for some people. If there is a large file size drop from 150 to 250 GOP, it means somebody is sitting still enough to win at the Mannequin Challenge.

[kagsundaram]

@Austin

it means somebody is sitting still enough to win at the Mannequin Challenge

I doubt that. I have always been fidgety, I talk with my hands, and now Parkinson’s is added to the mix.

[Austin]

Wide angle lens so all the movement is happening in a smaller square in the middle of the frame rather than a head filling the entire frame? Less effective change would be happening that way. All I can do is theorize.

Obviously, go with whatever settings work for you. Optimal settings and file size expectations are very specific to types of footage, including the variety of colors and the brightness of the lighting. (Dark compresses better than bright.)

[kagsundaram]

@Austin

Wide angle lens so all the movement is happening in a smaller square in the middle of the frame rather than a head filling the entire frame?

Yep.