Robot Mastering

Getting your music mastered is an expensive business. For professional musicians its perfectly understandable to pay a highly skilled audio engineer to do this work for you. However for the home producer it is just too expensive. The challenge: Can an automated system which employs the latest machine learning technologies do as good a job?

How it works

I am not going to give away the secret sauce here, but I will explain the mechanisms used to handle truly massive audio files.

First off we upload directly to AWS S3 (Amazon Web Services Simple Storage Solution) by signing the headers and payload. The browser passes information about the upload to our server which then signs it and returns it to the browser. Javascript in the browser can then initiate a direct upload to AWS. Once the upload is complete a job is added to a queue (also hosted by AWS). A pool of servers (which scales with demand) then takes jobs from the queue and processes the files. On completion the mastered file and a preview is saved back to another AWS S3 bucket and an email is sent to the user via Mailgun (an email service provider). The user can then preview the file and if it meets their expectations choose to pay for it via a payment gateway (Braintree). Once they have paid the mastered audio file can be downloaded.