Simulating TV noise

拟墨画扇 提交于 2019-11-28 03:36:10

Create video and audio noise, artifacts, and errors with ffmpeg

Noise

Using filters

The geq (video "generic equation") filter (with nullsrc as its "blank canvas") can create video noise, and the aevalsrc filter can create white noise audio:

ffmpeg -f lavfi -i nullsrc=s=1280x720 -filter_complex \
"geq=random(1)*255:128:128;aevalsrc=-2+random(0)" \
-t 5 output.mkv

Note that this will create black and white video noise.

Using /dev/urandom

Although I recommend using the geq filter you can also use /dev/urandom to generate video and audio noise (Windows users will have to use the geq filter as shown above). This is a small screeenshot, but of course the video will not be a static image:

ffmpeg -f rawvideo -video_size 1280x720 -pixel_format yuv420p -framerate 25 \
-i /dev/urandom -ar 48000 -ac 2 -f s16le -i /dev/urandom -codec:a copy \
-t 5 output.mkv

This will create color video noise. If you just want black and white you can add the hue filter.

ffmpeg -f rawvideo -video_size 1280x720 -pixel_format yuv420p -framerate 25 \
-i /dev/urandom -ar 48000 -ac 2 -f s16le -i /dev/urandom -codec:a copy \
-t 5 -vf hue=s=0 output.mkv

Adding random visual noise and errors to an existing video

Using the noise bitstream filter:

Original and modified versions.
ffmpeg -i input.mp4 -codec:v huffyuv -bsf:v noise -codec:a copy noise.mkv

According to the documentation:

A bitstream filter operates on the encoded stream data, and performs bitstream level modifications without performing decoding.

This bitstream filter can accept a value to increase or decrease the amount of noise. It's inverse, so a higher number is less noise, and 1 is the lowest number and therefore the most noise. You will need to experiment to see what works best for you.

This first example stream copied the audio and only glitched the video, but you can apply the filter to both audio and video by removing the stream specifier:

ffmpeg -i input.mp4 -codec:v huffyuv -c:a pcm_s16le -bsf noise=1000000 noise.mkv

Or provide separate values for video and audio:

ffmpeg -i input.mp4 -codec:v huffyuv -c:a pcm_s16le -bsf:v noise=1000000 -bsf:a noise=100 noise.mkv

This seems to work well with rawvideo or huffyuv for video, and pcm_s16le for audio, but I recommend experimenting. See lossless vs corruption (notes) for a video showing how different encoders react to noise corruption.

The output noise.mkv is damaged, but you can re-encode it so it will work in your player:

ffmpeg -i noise.mkv -codec:v libx264 -pix_fmt yuv420p output.mkv

See the H.264 and AAC encoding guides on the FFmpeg Wiki for more encoding info.

Macroblock effect

There are several methods to do this, but in this example the noise bitstream filter will be used. Output to MPEG-2 video in TS as this will be more susceptible to the desired effect:

ffmpeg -i input.mp4 -bsf:v noise -c:v mpeg2video -q:v 2 -c:a copy macroblock.ts

Increase -q:v value to ~20 if you want a more "encoded" look. Add a noise value as shown in the section above if you want more noise.

The output macroblock.ts is damaged, but you can re-encode it so it will work in your player:

ffmpeg -i macroblock.ts -codec:v libx264 -pix_fmt yuv420p output.mp4

See the section above for more info on the noise bitstream filter.

Results may vary, so you may not get what you're looking for. See hex editor method below.

Using a hex editor

Another method is to use a hex editor which you may find to be easier and more controllable and more consistent. See:


Forcing a pixel format

You can lie to ffmpeg and make it think that the colorspace and chroma subsampling is different that it actually is resulting in strange, error-like effects.

Original and modified versions.
  1. Probe your input.

    ffmpeg -i original.mp4
    

    Note the frame rate and video frame size.

  2. Refer to ffmpeg -pix_fmts and choose one of the available formats such as yuv420p16le.

  3. Create rawvideo and pipe it to another ffmpeg instance:

    ffmpeg -loglevel error -i original.mp4 -f rawvideo - | ffmpeg -y -f rawvideo -framerate 25 -video_size 1280x720 -pixel_format yuv420p16le -i - -pix_fmt yuv420p video.mp4
    

    The -framerate and -video_size values were copied from the original input file information shown in the console output of step 1. Of course you can also manipulate these for various effects.

  4. Mux audio if desired

    ffmpeg -i video.mp4 -i original.mp4 -map 0 -map 1:a -c copy output.mp4
    

yes, you can use ffmpeg:

ffmpeg -s 640x480 -t 8 \
-f rawvideo -pix_fmt gray16be -i /dev/urandom \
-f s32le -i /dev/urandom \
output.mp4

Just for comparison, I tested 3 different ways to generate pseudorandom numbers. As you can see the speed is very different. For high badwidth I suggest the first solution. You can find the source here: https://github.com/Zibri/rand

By the way, the source can be further optimized to gain ever better speeds.

$ ./rand|pv -S -s 2G >/dev/null
2,00GiB 0:00:01 [1,58GiB/s] [========================================================================>] 100%            

$ cat /dev/urandom|pv -S -s 2G >/dev/null
2,00GiB 0:00:11 [ 178MiB/s] [========================================================================>] 100%            

$ dd if=/dev/zero bs=1k count=2M 2>/dev/null | openssl enc -rc4-40 -pass pass:weak |pv -S -s 2G >/dev/null
2,00GiB 0:00:04 [ 413MiB/s] [========================================================================>] 100%            

it's white noise

create images in grayscale with a random intensity and audio with completely random samples

易学教程内所有资源均来自网络或用户发布的内容,如有违反法律规定的内容欢迎反馈
该文章没有解决你所遇到的问题?点击提问,说说你的问题,让更多的人一起探讨吧!