Creating video from images WITH audio

[boyer]

Hi guys and gals!

First off let me say, this is a great community! Long time reader, first time poster.

Ok, here's my question:

I have a series of sequentially named images (Slide01.jpg, Slide02.jpg, ...), and sequentially named wave files (Slide01.wav, Slide02.wav, ...).

What I'd like to do is build a video where the images are synced with the audio, so Slide01.jpg is displayed the whole time Slide01.wav is playing, then Slide02.jpg is displayed while Slide02.wav is playing, etc.

I have a directory which contains both the images and the wav files.

I've been monkeying around with ffmpeg, and can get it to build a video from the sequence of images w/o audio just find by doing:

ffmpeg -i Slide%02d.jpg -r 15 output.flv

The %02d means "for each file in this directory named Slidexx.jpg (where xx is a 2 digit number).
The -r switch sets the framerate

I've tried to add the audio in as a mapped stream with this line:

ffmpeg  -i Slide%02d.jpg -i Slide%02d.wav -map 0:0 -map 1:0 -r 15 output.flv

Seems like it should work, however, ffmpeg barfs telling me that it can't find "Slide%02d.wav".

I haven't been able to find any infos on the iwebs regarding adding audio to a video being created from images. I KNOW this has to be possible....

Any ideas gang? I'm open to using a tool other than ffmpeg, and don't care about platform as we do an equal amount of work in both Windows and Linux.

Thanks in advance,
Sean

[aedipuss]

i know dvdslideshowgui can sync multiple slides to one audio, don't know about many but it may be worth trying. as it's free anyway.

http://download.videohelp.com/tin2tin/

if you had no other choice you could use it to do each slide/song individually and then join all the resultant vobs into one dvd.

[boyer]

Thanks for the info!

The video I'm intending to create is not destined to become a DVD. It's going to ultimately be a .flv (flash video).

I had the same thought about creating separate clips for each audio/image pair, then joining, and am currently fighting with a Windows batch script to do this for me. I think I'll move over to my Linux box as Batch makes me want to rip my hair out by the handful!

I'd also like to stay in the command line as this process will need to be scripted.

Any other thoughts?

[boyer]

Ok, so after more playing around with things, what I've been doing is this:

I have a shell script that loops over a directory. Here's my script:

#!/bin/sh

for f in /home/boyer/VideoTest/*.wav

	do
	{
	  wav=`basename $f`
	  name=`basename $f .wav`
	  jpg="${name}.jpg"
	  output="${name}.avi"

	  convert $jpg -resize 656x492\! -quality 100% $jpg
	  ffmpeg -i $wav -i $jpg $output

	}
done;

for f in /home/boyer/VideoTest/*avi
	do
	{
	  file=`basename $f`
	  files="${files} $file"
	}
done;

mencoder -ovc copy -oac copy -o joined.avi $files

As you can see, for each file in the dir, I'm resising my image (from 660x495) to 656x492 (perfect 4:3 ratio). Then I'm running ffmpeg, giving it a wav source, and a jpg source.

An avi (I've also tested with mpg and flv) is outputted.

For whatever reason, the video files produced with ffmpeg are corrupt. The avi's have broken headers and can't seek and won't combine), and the mpgs won't play video at all...

Then I'm attempting to combine (in this case the avi's) using mencoder, but it tells me my source videos are broken, and thus my output video is broken.

It's clear that something is going wrong from the very beginning. Should I be including some -map options on my ffmpeg call?

Advice?

Thanks!
Sean

[Soopafresh]

Perhaps you should specify a different codec to use

-vcodec libx264 -b 2000

or use a lossless one like -vcodec huffyuv then convert the final product in mencoder to something that will work in an flv container (typically h264)

[Khaver]

Here's an avisynth script that will create a video of your image at your desired frame rate for the duration of the length of your wave file.

Just feed the AudioNImage function the location of the wave file, the location of the image file, and the required frames per second of the video.

Example:

audio="D:\video\Captures\rocky_mountains.wav"
picture="D:\video\Captures\firstframe.bmp"
AudioNImage(audio,picture,15)

#spline32resize(640,480)

function AudioNImage(string wave, string image, float fps)
{
aud=Wavsource(wave)
length=audiolengthF(aud)
rate=Audiorate(aud)
frames=Int((length/rate)*fps)
vid=ImageSource(image, end=Int(frames), fps=fps)
return audiodub(vid,aud)
}

You can add a resize after the function if needed.

You can then feed ffmpeg like this: ffmpeg -i avsfile.avs outputfile.flv

[boyer]

Cheers you guys! Saved me with the avisynth

I'd still love to figure out why I can't get a usable video out of ffmpeg on it's own, but this get's me going for now!

I'll post back if I ever discover a way to do it without avisynth.

Thanks again!

[Rotsujin]

You can then feed ffmpeg like this: ffmpeg -i avsfile.avs outputfile.flv

Hi I am trying to do the same thing, I got your script to work (I had to change the output file extension to .avi), but the image in the output video is flipped vertically.

Do you know how to fix that?

[boyer]

Oddly enough the same thing was happening to me too!

All I did was flip my images before hand using image magick. This was ok, since I was resizing the images explicitly first anyway. Here's my line:

imgconv image.png -resize 656x492! -flip imagesmall.png

[Gavino]

If necessary, you could also do the flipping in the Avisynth script by changing
vid=ImageSource(image, end=Int(frames), fps=fps)
to
vid=ImageSource(image, end=Int(frames), fps=fps).FlipVertical()

[Rotsujin]

@boyer & Gavino: Thanks for the help!

[boyer]

Hey guys!

Back again with another question regarding this issue.

I've had to abandon using avisynth, as for the live of me (and our IT team) we cannot get avisynth to build on Linux (which is ultimately where this stuff has to run).

Using ffmpeg, I can create a video from a png image, and a wave file.

The problem is, that the video stream is only 34ms long where the audio stream is 4:43 let's say.
On Windows, this avi will play/seek just fine, but on Linux, it is totally broken.

My question is, is there a way to sync the length of the video stream (in this case the input is a png image) to the length of the audio stream, without knowing the length of the wav at runtime?

If worse comes to worse, I could find a tool to extract the runtime of the wav, then manually put that into the ffmpeg command, but I'd rather avoid that if possible.

Any ideas?

[boyer]

PS - the solution for creating a piece of video from a still image with audio is using the -loop_input switch on the image input side.

ffmpeg -loop_input -r 1 -i image.png -i audio.mp3 output.avi

The above code will loop the single image the entire length of the mp3 audio file given. The video stream will be VERY small, because you can input at 1 frame per second. If your output format won't support 1 fps, just jack it up on the output. Voila!

[bseos]

*del