Everyone's raving about the quality, but no one mentions the operational overhead. Downloaded a batch of 20 tracks, and the LUFS range was insane. Had to build a post-processing step just to make them usable.
My normalization workflow (FFmpeg, because paying for another cloud service to fix this is ironic):
```bash
# Analyze loudness
ffmpeg -i input.wav -af loudnorm=I=-16:TP=-1.5:LRA=11:print_format=json -f null -
# Apply normalization to -16 LUFS
ffmpeg -i input.wav -af loudnorm=I=-16:TP=-1.5:LRA=11 output_normalized.wav
```
Cost breakdown for processing 1000 tracks:
* Spot instance (c6a.2xlarge) time: ~0.5 hours
* Estimated cost: $0.15
* Versus manual leveling in a DAW: priceless (and not in a good way)
show the math
show the math
Your cost breakdown is off. You're not factoring in development time to automate that batch process, or maintenance when FFmpeg updates break your loudnorm flags.
I ran similar numbers last month. Actual total for 1000 files was closer to $0.45 once you account for the initial analysis pass and file transfer overhead on the instance. Still cheap, but the real cost is the pipeline itself. What happens when you need to process 20,000 and your spot instance gets reclaimed mid-job?
Your CRM is lying to you.