03 Sep
03Sep

Medium

LinkedIn

Nine months ago, I began a quest to combine the art of audio drama creation with the latest AI technology. The result is Son of Yi, Son of Pangu, a three-hour audio epic that debuted on YouTube and X on Labor Day 2024.

This AI radio play was completed in mid-August after production began in January of this year. It is the biggest creative project I have ever undertaken, my first original radio play, and a major milestone in the brief but exciting history of AI in storytelling.

I invite you to listen to Son of Yi, Son of Pangu and to share your thoughts in the comments section.

Background

I had previously used AI software to produce a Batman fan audio drama. But I was inspired to create an original, feature-length AI radio play after a fellow AI enthusiast encouraged me to make up a story of my own. Thus, my quest to make an original AI radio play began.

My goal when creating Son of Yi, Son of Pangu was to do something new with the fantasy genre, and also to stretch the limits of the audio drama/radio play medium. I saw that the market seemed to be saturated with fantasy literature, and that a lot of contemporary fantasy literature seemed to be very homogenous and derivative. So I hypothesized that one way to do something new in the fantasy genre would be to try telling a fantasy story in a different medium. This led me to dive into audio drama production and to begin work on Son of Yi, Son of Pangu.

My Process

1. Tools

The main AI tools I used were:

  • Play: An AI voice synthesizer.
  • Aiva: An AI music generator.
  • Suno: An AI song generator.
  • Zapsplat: Source of royalty-free sound effects.
  • Audacity: Open-source audio editing software.

I used Play to generate the character voices in my audio drama because it was easy-to-use, affordable, and had the best synthetic voice-acting I could find. Many people suggested that I use ElevenLabs, but I was already in the middle of the project by the time I learned about it.

I used Aiva to generate the incidental music for my radio play, with it again being easy-to-use and affordable.

I used Suno to generate the vocals for the songs in my radio play (I wrote all of the lyrics with no assistance from AI) because it was the first AI song generator I became aware of and because it quickly proved reliable and affordable. I had also heard of Udio, but I was so pleased with the results Suno provided that I did not explore using Udio. I also used Suno to generate parts of my radio play’s musical score.

I got sound effects from Zapsplat owing to the comprehensive, inexhaustible supply of high-quality sound effects it offered. Zapsplat’s affordable, royalty-free offerings were very attractive, with an easy-to-use website.

I used Audacity because it was open-source and very user-friendly, and was designed for beginners and hobbyists, making Audacity very attractive to an amateur audio engineer such as myself.

2. Innovations

After I selected my tools and had written a script, I began producing the radio play. I used Play to generate the lines of dialogue, one-by-one, clicking-and-dragging the generated audio into the Audacity files. When I needed incidental music, I generated it using Aiva or Suno, and did the same. When I needed a sound effect, I searched for the appropriate sound in Zapsplat, and put the audio file for it in the Audacity file, and edited it as needed.

The main innovations which I made in this project were:

  1. Producing an original radio play script which was not composed with AI.
  2. Using Play, with its text-to-speech function, to create customized generations of the script’s dialogue.
  3. Using Aiva and Suno to create AI-generated music and songs, making edits to the generated audio as needed within Audacity.
  4. Creating a 3-hour supercut, which I imagine is certainly one of the longest original, single audio dramas in existence, and certainly the longest one made using AI.

3. Challenges Faced

The biggest challenges when creating Son of Yi, Son of Pangu were:

  1. Finding AI voice-clones for the main characters who were good actors.
  2. Figuring out creative ways to work with the stock sound effects to create the desired sounds.
  3. Teaching myself how to use Audacity effectively.
  4. Managing my time and schedule so that the project would be completed efficiently.
  5. Communicating the nature of the project to others and explaining what it was. Many people, whether they were AI enthusiasts or not, were not even sure what “a radio play” even was, or what I meant when I said I was “using AI” to make it.

However, I was able to overcome these obstacles and complete the project.

4. Solutions

Through trial-and-error, plus the critical advice of some friends, I found AI voice-clones within Play’s system which were either excellent voice-actors or at least were good enough.

I figured out ways to manipulate stock sound effects from Zapsplat to produce desired effects, and sometimes accidentally stumbled upon sounds which were even better than what I originally had in mind.

I worked with Audacity, despite my lack of form training in how to use it, and adapted to gaps in my knowledge by looking up information on the internet about how to use the software.

When I began to get tired and disorganized, I took a break from producing. When I returned to work on the radio play, I made a schedule of the work I wanted to get done, decided on a deadline, and rationed out what work had to be done each day and each week for fifteen weeks in order to finish the project by the deadline I had set. Following this system, I was able to work efficiently, avoid burnout, and complete the whole project about two weeks ahead of schedule.

I still have some difficulty explaining what I am doing to people, either within the AI community or beyond it. The reality is that the radio play or audio drama is an obscure, niche art form which not many people have extensive knowledge about. A lot of people, when I tell them about Son of Yi, Son of Pangu, think I am talking about making an audiobook, which is different from an audio drama or radio play.

My solution to this problem is to focus on showing my radio play to audio enthusiasts and radio professionals who do have some knowledge of the nature of radio plays and audio dramas. I find that such people are more likely to understand and appreciate my work, and perhaps help me to do better in sharing it with a wider audience.

The Final Product

Son of Yi, Son of Pangu is a high fantasy adventure epic about a warrior named Bo and a talking phoenix named Lo, who live in a mystical country trapped in darkness and which has no sun (the sun disappeared after the Sun God was kidnapped and imprisoned by an evil god named Gong, who is also the Demon King). After Bo and Lo meet, they go on a quest to defeat the Demon King and rescue the Sun God, knowing that if they succeed, light, prosperity, and happiness will be restored to their country.

The two main characters are a young warrior named Bo, who is half-god, half-man, and a talking phoenix named Lo, who is able to use powerful magic and who secretly wishes to become a human being. Bo, a brave young man, and Lo, who is a wise, female spirit as well as a talking bird, become best friends during their quest, and they learn how to trust and respect each other’s strengths and abilities.

The main theme I wanted to communicate in this story is that friendship is instrumental in overcoming evil. I believe that evil really does exist in the world, but that the best way to fight it is simply to be good, kind, and loving to other people in your everyday life, and to stick up for your friends and family during hard times. That, I believe, will do a lot more good than all the laws, policies, organizing and activism in the world.

AI played no part in the composition of the script for Son of Yi, Son of Pangu, but the interpretations of the dialogue which the AI voice-clones which I was using to bring the characters to life played a major role in shaping the tone of the finished story. Live voice-actors probably would have interpreted the characters included in the script differently, and a live cast may have pushed back on some parts of the script and perhaps have argued for making story or dialogue changes in the middle of the production. AI voice-clones had no such complaints, but working with Play’s system to produce convincing dialogue was still a challenging and fascinating process.

My use of AI-generated music was also a crucial component of the production. I included songs in Son of Yi, Son of Pangu, in a manner similar to a classic Disney film. These songs are generated by Suno, and I originally wanted the songs to all be classical and operatic in their style. But Suno generated a lot of versions of the songs which sounded much more modern, with a lot of rock, techno, and electronic elements in these songs. These versions of the songs really grew on me, leading me to include a much more eclectic mix of musical styles in the radio play.

Listen Now

Son of Yi, Son of Pangu can be listened to either on YouTube or on Spotify. It is divided into five parts on those platforms, with links embedded below. Note that the 3-hour supercut is only available on YouTube.

I invite all listeners to share their thoughts on the radio play with me in the YouTube comments section, or the comments section of this post. Feel free to reach out to me by email or direct message as well.

Behind the Scenes

I plan on creating a YouTube video recording me using all the techniques I used to create Son of Yi, Son of Pangu to produce a short AI radio play script, where I will use screen capture to record the production process while giving running commentary on that process. This behind-the-scenes video will be out on October 7, 2024.

I am currently considering what future projects I want to create in AI audio. I am already writing new scripts for new AI radio plays, and I am contemplating using the techniques I used to make Son of Yi, Son of Pangu to produce dramatic readings of short stories and poems I will write.

Conclusion

I would like to thank everyone who supported me while I undertook this project, especially the online generative AI community on X, as well as the members of the Seattle AI Society. Son of Yi, Son of Pangu is dedicated to the memory of my late grandfather, Myrol Allen “Pat” Sweeney, who passed away in 2023.

I happily invite you to listen to my radio play, share it with others, and also to subscribe to my personal website and follow me on X.

Leave a comment sharing your thoughts on Son of Yi, Son of Pangu, and if you have any questions or comments you'd like to send Levi directly, reach out to him by email at info@bolostudios.io

Comments
* The email will not be published on the website.