AI in Action: NotebookLM to Video Podcast in 30 Minutes

AI has revolutionized the world of content creation, making it easier and faster to produce high-quality videos. As a recent example, I created a video podcast in 30 minutes using a variety of AI tools (NotebookLM, CapCut, ElevenLabs, Canva, & GoogleLab's MusicFX)

Here's the result:

Here's a step-by-step guide on how I did it:

1. Creating Unique Voices with AI

I started with an audio podcast that my colleague, Vivek Bapat, created using Google's new NotebookLM tool. While the voices generated by NotebookLM are impressive and realistic, I wanted to give the podcast a more unique feel.To achieve this, I used Eleven Labs' AI-powered Voice Changer to modify the voices. This involved a two-step process:

Splitting the Audio: First, I split the two voices into separate audio files using the AI-enhanced Capcut editing tool. This involved manually selecting and separating the audio tracks for each speaker, which was the most time-consuming part of the process.

Transforming the Voices: Next, I uploaded the separated audio files into Eleven Labs' Voice Changer. I chose two new voices that were distinct from the originals and let the application generate modified audio files.

2. Bringing the Voices to Life with AI Avatars

With my unique audio tracks ready, it was time to create the video component of the podcast. I used Hey Gen's Avatar creation suite to generate video hosts. I selected two realistic avatars and customized their poses and backgrounds to fit the podcast's theme. For each avatar, I created a separate project, uploaded the corresponding audio file, and generated a video of the avatar speaking.

3. Polishing the Production: Editing and Enhancing with AI

Now that I had my "presenters," I used CapCut to assemble the video podcast into a more professional format. I positioned and cropped the two video files side-by-side. To minimize distracting hand gestures from the male avatar, I zoomed in on his image.

4. Adding the Finishing Touches: Intro, Music, and Export

To create an intro for the video podcast, I used Canva Pro to design a simple template, which I then imported into CapCut. I also generated two additional intro videos using HeyGen, featuring different angles of the avatars, to simulate footage from another episode.

Finally, I used Google Lab's MusicFX tool to generate a background audio track. I chose a "mid-tempo ambient theme for technology podcast" and exported a 30-second looping track. In CapCut, I added this track and lowered the volume after the intro segment to create a subtle audio bed.

The entire process took me approximately 50 minutes, with the audio separation taking the most time (around 25 minutes). I anticipate this could be done in under 30 minutes once an AI tool for audio separation becomes available.

Key Takeaways:

• AI makes it incredibly easy and fast to create professional-quality video podcasts.

• Creativity still matters. Simply using these tools "as is" will produce generic content at scale that doesn't resonate with audiences.

• It is important to be mindful of the potential for AI-generated content to appear synthetic and to take steps to avoid this.

Conclusion

Although this AI-generated video podcast produced amazing results in a short amount of time, it's ultimately a thought exercise to examine the current "art of the possible." I believe it's only a matter of time (months? days?) before a company like Google or OpenAI builds this functionality directly into their products and reduces the time to creation from 30 minutes to 30 seconds. When that happens, we'll need to explore how to use these new capabilities in interesting ways to produce even more authentic content that resonates with our audiences.

This rapid advancement in AI technology highlights the need for continuous learning and adaptation in the content creation space. By staying informed and experimenting with new tools, we can harness the power of AI to enhance our creativity and productivity.