Testing AI voice overs.
The first is elevenlabs and of course it’s much better in quality and inflection. https://georgesaoulidis.com/elevenlabs
But the second one is open source running locally on my DGX Spark. It’s also much shorter.
So I added them to the pipeline for @9godsorg and even through the elevenlabs API is quite cheap for a minute of audio, I want to cut down on subscriptions and use local models.
The generations are fast enough to run Cron jobs and make them in small batches every day, seems to work just fine for now.

0 Comments