React Native Radio Fake News
An AI made this Podcast Episode
If you haven’t seen it yet, I made a fictitious React Native Radio podcast episode for Infinite Red in record time using a new AI tool.
While React Native Radio is always known for being fun and funny, a lot of work goes into every episode… except the one above.
How long did it take?
I had the whole thing done in about 30 minutes from idea to solution. I guess I spent more time on this blog post than I did doing the thing.
A year ago, I did something similar for a joke on Twitter, and it took longer and was a lot harder. It also didn’t let me pick out who to use as a voice!
I’m amazed at how easy it was today.
Why do this?
Besides being a fun example, this tooling could be used to help you handle long blog articles in a friendly voice, filling in gaps in podcasts, and even creating an amazing presentation at the last moment, without haveing the talent run to a recording booth!
This is a new AI future, and it’s easy.
How did I do this?
First, I needed to get over a minute of audio from our hosts. That’s not hard, because they upload their individual audio files after each recording of https://reactnativeradio.com/
Jumping into the thread, I grabbed 40ish minutes of their audio.
The audio is in the lossless .flac format, so I had to quickly convert it to mp3. I then broke it into chunks of about 10MB mp3s.
Next, there are a lot of open source repos to make text to audio, but I’ve been significantly impressed with the work of the Eleven Labs product which has a free tier: https://beta.elevenlabs.io/
From their service, I used their “Voice Cloning”
Here I created new “Instant Voice” for my podcast guests, and uploaded their samples for training.
After the models were ready, it was just a matter of writing an award winning script and then generating each person’s parts.
At first, the voices were pretty flat, but once I opened the voice settings, I brought the stability down to 25% and bumped the clarity/similarity up to near 90%.
et voila! It was extremely accurate with some voices, but only slightly accurate with others.
Once I had all the generated dialogue set, I just brought it into an editor to cut and smash it back together in conversational format.
The result is a podcast that never really happened!
Here at Infinite Red, we’ve been working on fun but useful AI projects for years. If you’re looking for React Native consulting or want to talk about integrating AI into your mobile app, reach out to us!
Gant Laborde is a co-owner and Chief Innovation Officer at Infinite Red, published author, adjunct professor, worldwide public speaker, and mad scientist in training. Clap/follow/tweet or visit him at a conference.
