Today I tried out an open-source voice cloning project just to see how good the results would be.
Deploying this open-source project turned out to be pretty simple — it runs fine on a consumer-grade GPU. The voice cloning results were better than I expected going in.
Saturday morning, I went ahead and wired this feature together with my document-editing pipeline, so that a document with text and images gets automatically turned into a video. That means whenever I publish a new blog post, a short video gets generated for it automatically at the same time.
There’s obviously still some rough edges to polish, but I’m treating that as part of the fun.
