CODEWITHSUNDEEP
amazon-web-servicesamazon-s3ffmpeg
Dario Rusignuolo
Dario Rusignuolo

Reputation: 2186

Read S3 video file, process it with ffmpeg and upload to S3

I have a video stored in s3 bucket with authenticated-read ACL.

I need to read and make a trailer with ffmpeg (nodejs)

Here's the code I use to generate the trailer

exports.generatePreview = (req, res) => {
    const getParams = {
        Bucket: S3_CREDENTIALS.bucketName,
        Key: req.params.key
    }
    s3.getSignedUrl('getObject', getParams, (err, signedRequest) => {
        console.log(signedRequest, err, 'getSignedUrl')
        ffmpeg(new URL(signedRequest))
            .size('640x?')
            .aspect('4:3')
        .seekInput('3:00')
        .duration('0:30')
        .then(function (video) {
            s3.putObject({ Bucket: S3_CREDENTIALS.bucketName, key: 'preview_' + req.body.key, Body: video }, function (err, data) {
                console.log(err, data)
            })
        });
});

}

Unfortunately, the constructor path seems not to read remote url. If I try to execute an ffmpeg command line with the same signedurl (i.e. ffmpeg -i "https://[bucketname].s3.eu-west-1.amazonaws.com/[key.mp4]?[signedParams]" -vn -acodec pcm_s16le -ar 44100 -ac 2 video.wav)

The error I get is that the signedRequest url 'The input file does not exist'

It seems fs.readFileSync https is not supported even if I try the request with http with the same result. fs.readFileSync(signedurl) => gives the same result

How to overcome this issue?

Upvotes: 1

Views: 4383

Answers (1)

Brian Burton
Brian Burton

Reputation: 3842

If you're using node-ffmpeg this isn't possible because the library only accepts a string pointing to a local path, but fluent-ffmpeg does support readstreams so give that a try.

For example (untested, just spitballing):

const ffmpeg = require('fluent-ffmpeg');
const stream = require('stream');

exports.generatePreview = (req, res) => {
  let params = {Bucket: S3_CREDENTIALS.bucketName, Key: req.params.key};

  // Retrieve object stream
  let readStream = s3.getObject(params).createReadStream();

  // Set up the ffmpeg process
  let ffmpegProcess = new ffmpeg(readStream)
    //Add your args here
    .toFormat('mp4');  

  ffmpegProcess.on('error', (err, stdout, stderr) => {
    // Handle errors here
  }).on('end', () => {
    // Processing is complete
  }).pipe(() => {
    // Create a new stream
    let pt = new stream.PassThrough();

    // Reuse the same params object and set the Body to the stream
    params.Key = 'preview_' + req.body.key;
    params.Body = pt;

    // Upload and wait for the result
    s3.upload(params, (err, data) => {
      if (err)
        return console.error(err);

      console.log("done");
    })
  });
});

This will have high memory requirements so if this is a Lambda function you might play around with retrieving only the first X bytes of the file and converting only that.

Upvotes: 3

Related Questions