image + audio = mp4: how to make video smaller?

Question

I'm using ffmpeg to add a JPG to an MP4 to make an MP4.

As there is only one image to make the video I'm wondering if there is a way to reduce the video size?

Command:

$ ffmpeg -loop 1 -i image.jpg -i audio.mp3 -shortest -c:a copy output.mp4

Results:

• image.jpg = 26.7K (image not so clear)
• audio.mp3 = 64.6M (54 minutes)
• output.mp4 = 80.6M (video result is not so clear, music still good)

Is it a reasonable size for MP4?

Answer 1

For uploading to YouTube

H.264: smallest files

This method uses libx264 to encode H.264 video. It is slower than the stream copy method below, but potentially will output a smaller file size.

ffmpeg -loop 1 -framerate 1 -i image.jpg -i music.mp3 \
-c:v libx264 -preset veryslow -crf 0 -c:a copy -shortest output.mkv

Stream copy: fastest process

This method just stream copies (no encoding) the image into the MKV container. It's super fast, but if size is important than the method above may produce a smaller file.

ffmpeg -loop 1 -framerate 1 -i image.jpg -i music.mp3 -c copy -shortest output.mkv

• YouTube accepts just about anything so these commands use a few tricks to make encoding faster, or make the file size small, and to keep quality high because YouTube will re-encode whatever you give it. Your player probably won't like it but YouTube will.

• In these examples a very low frame rate is used which should process faster than the default of 25 fps.

---

Widest compatibility for any player

ffmpeg -loop 1 -i image.png -i music.mp3 -vf "scale='min(1280,iw)':-2,format=yuv420p" \
-c:v libx264 -preset medium -profile:v main -c:a aac -shortest -movflags +faststart output.mp4

• This should play on just about anything except very ancient devices (change -profile:v main to -profile:v baseline if that is the case).

• If your audio input is already AAC then change -c:a aac to -c:a copy to avoid unnecessary re-encoding.

• Encoding time will be longer and file size will be bigger than the YouTube commands above.

• See FFmpeg Wiki: H.264 for more info.

Answer 2

With only the information you provided this what I can think of.

Let's go by pieces:

As there is only one image added to the video. I m wondering if there is a way to reduce the video size eg force bit rate/second = 1 instead of 28 ?

As it is a still image, no need to change. just stay still.

What you're talking about there it's frames per second (fps) not bitrate. And you're right in the concept. Beign just one still image 1 fps wouldn't be a problem.

Is it a reasonable size for mp4? Does image size matter a lot?

In this case image size wouldn't have a significant weight because we are dealing with just one still picture. If it were a high quality video then it's a different issue.

Should I resize jpg to 1080 * 720, and make additional settings in ffmpeg to output as 720p (what is the ffmpeg setting then :) ?

The resolution you are saying it's 1280x720.

I dont' see the point in the -loop 1 option here, it will only slow down the process. Did you use -shortest cause the file has more than 1 input stream? If no then there's no reason to use it. Since you used -acodec copy I will have to assume that you don't want to re-encode it.

For an output of 720p 1fps without re-encoding and not touching bitrate:

ffmpeg -framerate 1 -i input.mp3 -i cover.jpg -c:a copy -s 1280x720 output.mp4

Beign the case that it has more than one input stream and you want ffmpeg to finish encoding at the shortest stream then:

ffmpeg -framerate 1 -i input.mp3 -i cover.jpg -c:a copy -s 1280x720 -shortest output.mp4

Now if you have to change the bitrate (let's say 1024k as an example);

ffmpeg -framerate 1 -i input.mp3 -i cover.jpg -c:a copy -b:v 1024k -bufsize 1024k -s 1280x720 output.mp4

See: Limit the output bitrate

-c:a copy = Copy the input audio stream so we not re-encode.

-b:v = The bitrate specified to be (not precisely) constant.

-bufsize = It's the interval in which calculates the average bitrate. Meaning the lowest more loss quality image but more accurate bitrates according to the specified in -b:v.

-s = The actual size of the video output.

-framerate = The input frame rate we want that the output will use.

Note that if you still want a smaller size video at one point you will have to sacrifice quality in order to achieve that. And other ways to do it involve re-encoding with lossy compression.

From what you say, the information you provide and what you are trying to do I think it seems that your target is to embed album-art to mp3. If that's your goal this would fit your requirements better:

ffmpeg -i input.mp3 -i cover.jpg -map_metadata 0 -map 0 -map 1 output.mp3

-map_metadata = The ffmpeg metadata specifier to output file from infile. In this case (zero) the global metadata.

-map 0 = Input stream 1 (audio).

-map 1 = Input stream 2 (image).

If nothing here suits your requests please give more information.