Upload FFmpeg output directly to Amazon S3

Question

I am using the fluent-ffmpeg library with node.js to transcode videos originally in a flash movie format to the mp3 format with multiple resolutions, 1080p, etc.. Once the transcoding is complete, I would like to move the transcoded video to an s3 bucket.

I pull the original .flv file from a source s3 bucket and pass the stream to the ffmpeg constructor function. The issue is after the transcoding completes, how do I then get the stream of the mp4 data to send to s3.

Here is the code I have so far:

        var params = {
            Bucket: process.env.SOURCE_BUCKET,
            Key: fileName
        };
        s3.getObject(params, function(err, data) {
            if (err) console.log(err, err.stack); // an error occurred

            var format = ffmpeg(data)
            .size('854x480')
            .videoCodec('libx264')
            .format('flv')
            .toFormat('mp4');
            .on('end', function () {
                //Ideally, I would like to do the uploading here

                var params = {
                   Body: //{This is my confusion, how do I get the stream to add here?},
                   Bucket: process.env.TRANSCODED_BUCKET,
                   Key: fileName
                };
                s3.putObject(params, function (err, data) {

               });
            })
            .on('error', function (err) {
                console.log('an error happened: ' + err.message);
            });

        });

For the code above, where can I get the transcoded stream to add to the "Body" property of the params object?

Update:

Here is a revision of what I am trying to do:

var outputStream: MemoryStream = new MemoryStream();

        var proc = ffmpeg(currentStream)
            .size('1920x1080')
            .videoCodec('libx264')
            .format('avi')
            .toFormat('mp4')
            .output(outputStream)
            // setup event handlers
            .on('end', function () {
                uploadFile(outputStream, "").then(function(){
                    resolve();
                })
            })
            .on('error', function (err) {
                console.log('an error happened: ' + err.message);
            });

I would like to avoid copying the file to the local filesystem from s3, rather I would prefer to process the file in memory and upload back to s3 when finished. Would fluent-ffmpeg allow this scenario?

Answer 1

You don't seem to be saving the output of the transcoding anywhere.

1. Save the output of the transcoding (your new .flv file) using output to your local filesystem.

2. Provide your new file's contents to putObject. According to the putObject documentation, the Body parameter accepts:

   Body — (Buffer, Typed Array, Blob, String, ReadableStream) Object data.

Here's some revised sample code:

// Generate a filename for the `.flv` version
var flvFileName = fileName.substring(0, fileName.length - path.extname(fileName).length) + '.flv';

// Perform transcoding, save new video to new file name
var format = ffmpeg(data)
    .size('854x480')
    .videoCodec('libx264')
    .format('flv')
    .toFormat('mp4');
    .output(flvFileName)
    .on('end', function () {
        // Provide `ReadableStream` of new video as `Body` for `pubObject`
        var params = {
             Body: fs.createReadStream(flvFileName)
             Bucket: process.env.TRANSCODED_BUCKET,
             Key: flvFileName
        };

        s3.putObject(params, function (err, data) {

        });
    })

Note: You may be able to create an output stream from fluent-ffmpeg and upload that stream to AWS S3, if you prefer, but this will complicate the logic and error handling.