From 330cd30dd269635f7a1eacb1bba236f50d4058db Mon Sep 17 00:00:00 2001 From: Cassidy Williams Date: Sat, 6 Jan 2024 22:46:32 -0600 Subject: [PATCH] AI voice test --- src/posts/ai-voice-test.md | 26 ++++++++++++++++++++++++++ 1 file changed, 26 insertions(+) create mode 100644 src/posts/ai-voice-test.md diff --git a/src/posts/ai-voice-test.md b/src/posts/ai-voice-test.md new file mode 100644 index 0000000..a656e5a --- /dev/null +++ b/src/posts/ai-voice-test.md @@ -0,0 +1,26 @@ +--- +layout: "../layouts/BlogPost.astro" +title: "Experimenting with AI voice" +slug: ai-voice-test +description: "I tried out the ElevenLabs Beta!" +added: "Jan 06 2024" +tags: [learning] +--- + +I tried out the [ElevenLabs](https://elevenlabs.io/) beta in February last year, and I finally uploaded the audio of my little test! + +In my experiment, I used [one of my older blog posts](https://blog.cassidoo.co/post/private-wins/). After passing in some voice samples to ElevenLabs, I had it read the blog post in "my" voice. Then, I recorded my own reading of the blog post, too. + +So now, you can hear the synthesized version and the real version back-to-back! + +**AI version:** + + + +**Real voice version:** + + + +The AI version is definitely pretty good. I think if someone didn't know my voice super well, they wouldn't really question it. I do think there's more... "life" to the real one, but maybe that's my own bias. In the AI one, you hear little blips where the AI "jumps" between sounds, and in my real voice, you can hear me stumble over my words a bit. What's particularly interesting to me is that the cadence is pretty spot on; both audios are exactly 1 minute and 19 seconds. + +Anyway, cool test, and I'd be down to experiment with this more in the future!