Historical case study. The project closed in 2026; this description covers its development period.

Audio transcoding pipeline

Server-side pipeline: FLAC/MP3/OGG upload, transcoding to multiple formats, S3 streaming. Lossless playback in the browser from any device.

Audio transcoding pipeline

Challenge

Meloteka is a cloud music service - upload your collection, listen from anywhere. People bring files in all formats, from FLAC at 40-60 MB per track to MP3 and OGG. I needed to keep the originals intact, generate streaming versions for the browser, and make playback start instantly. Early-stage startup budget.

Technical details

Solution

I built an async pipeline on Node.js. On upload, the file goes straight to S3 on Yandex Cloud - original untouched. A queue task is created in parallel, a worker picks it up and runs it through ffmpeg: AAC for browser streaming (128/256/320 kbps by tier), metadata extraction, waveform preview. Streaming via presigned URLs - the client gets a direct S3 link, CDN caches popular tracks. Lossless served through a separate endpoint with Range request support.

Result

Average processing time is 15 seconds per track (40 MB FLAC). Playback starts instantly thanks to CDN. Storage costs ~$0.005/GB per month. The pipeline handles up to 100 tracks in parallel.

194 views