Lower quality no difference in file size

Hi, we are using AWS Lambda@Edge to compress and resize images on the fly, I saw that the compression for jpeg’s wasn’t no where near fully compressed by default so I added the following options. requiredFormat can be either webp or jpeg.

I tested with quality 56 first and then 25 but it made no difference in file size for an original image that is 30 MB (6000x600 px), resized down to 1920 px. Both files were 861 KB…

const Sharp = require('sharp');
// ...
S3.getObject(s3Params).promise()
  // perform the resize operation
  .then(data => Sharp(data.Body)
    .resize({
      width: width,
      withoutEnlargement: true,
      quality: 25,
      progressive: false, // jpeg option
      optimiseScans: false, // jpeg, required mozjpeg support
      chromaSubsampling: '4:4:4', // jpeg
      trellisQuantisation: true, // jpeg, required mozjpeg support
      overshootDeringing: true, // jpeg, required mozjpeg support
      // lossless: true, // webp option
    })
    .toFormat(requiredFormat)
    .toBuffer()
  )
.then(buffer => {
  S3.putObject({
    Body: buffer,
    Bucket: s3_bucket,
    ContentType: contentType,
    CacheControl: 'private, max-age=31536000',
    Key: key,
    StorageClass: 'STANDARD',
  })

According to https://tinyjpg.com/ it can compress the image 45% further and the image afterwards looks perfect…

Some options state in the documentation that they require mozjpeg, how do I know if it’s installed on Lambda@Edge?

Using:

"sharp": {
  "version": "0.21.3",

[lovell]

You’ll need to compile libvips from source with support for mozjpeg and make it available globally. How you do that with Lambda@Edge is a question for AWS to answer or perhaps someone on StackOverflow to help with - good luck!

[OZZlE]

Seems like they already support mozjpeg already because the images becomes much smaller using Lambda than locally where I don’t think I have it installed 😃

(but don’t take my word for it)