How I Built an AI Avatar and Hype Video in 15 Minutes with Google Flow

Welcome back to How I AI! In this episode, I did something a little different, and frankly, I wasn't sure it was going to work. I decided to create a video avatar of myself using Google Flow and the new Gemini Omni video generation model. The goal? To go from zero to a full, one-minute hype video for this very podcast in about 15 minutes, starring an AI version of me.

I’m a product person and I love building with AI, but I am not 'video creative.' I would never be able to solo-produce a promotional video for my podcast—I wouldn't know where to start with brainstorming, framing, or editing. This experiment was all about seeing if these new multimodal tools can truly unlock creative abilities for people like me. Could I actually generate something usable with practically no knowledge of the tool or the process?

The whole thing was an adventure, complete with some hilarious and spooky results. I'm going to walk you through the entire process, from creating my digital twin to directing my first AI-generated video. Let's get into how it went.

Workflow 1: Creating Your Digital Twin with Google Flow

Before you can make a video with a consistent character, you need the character. The first step was to use Flow’s avatar creation feature to generate a digital version of myself. I’d tried this when the feature first came out and it failed, but I was determined to give it another shot.

The Step-by-Step Avatar Process

1. Initiate Avatar Creation: Inside the Google Flow interface, I selected the option to 'Create an avatar' and clicked 'Get Started'.

1. Scan and Connect: The app generated a QR code on my desktop. I scanned it with my phone to link the devices and begin the photo capture process.

1. Capture Your Likeness: My phone's camera opened, and the app prompted me to look at a series of numbers that appeared on my screen. It was a quick sequence: 17, 81, 49, 20, 25, 22. After that, it had me turn my head to the left and then to the right, giving me a checkmark for each successful angle.
2. Processing: Once the capture was complete, the app needed a couple of minutes to process the images and build the avatar. The message just said 'we're done' and I had to wait and see if it worked this time.
3. Meet Your Avatar: Success! A few minutes later, my avatar appeared. I have to say, it was a slightly strange, fish-eye lens version of me, but it was definitely recognizable. Now I was ready to use it to create a video.

Workflow 2: Directing an AI-Generated Hype Video

With my 'me' character ready, it was time to produce the hype video. What's interesting about Google Flow is that it's pitched as a full creative suite, not just a video generator. I wanted to see if it could act as my AI producer and help me brainstorm the entire concept.

Step 1: Collaborating with the AI Creative Director

I started by prompting the chat agent within Flow to help me create a storyboard. I wanted it to be a collaborative process.

My first prompt was:

help me create a storyboard for a hype video. For the How I AI podcast, I already have a character.

Named me, we can reference, help me come up with the few scenes that would make this great. This is a podcast by Claire about the best ways to use AI at work and in life.!

I loved that the AI didn't just spit out a storyboard. It asked me clarifying questions to understand the vibe I was going for: "Is she in a modern studio or perhaps a bright airy home office? Should it feel high tech and sleek or more grounded and lifestyle focused?"

This is exactly what a real creative collaborator would do. I gave it more direction:

she is in a dark home office, dark green walls. With books about AI and fun posters, lighting around. This should be more authentic lifestyle version, but it's high tech and about coding. Have a hacker vibe to it.

Based on that, it generated a full, seven-scene storyboard. It included everything from an extreme close-up of my hands on a mechanical keyboard to a wide shot of the office, and even a cheesy-but-fun moment with a digital heads-up display. The AI had a full narrative arc planned out.

Step 2: Generating the Scenes with the Avatar

Now it was time for the magic. I took the scene descriptions from the storyboard and used them as prompts for the video generation model. A key step was to replace any mention of 'Claire' with the @me tag to reference my newly created avatar.

I did make one rookie mistake right away—I accidentally generated images instead of videos because I didn't have the right mode selected. It's a simple toggle at the bottom of the input box, but an easy thing to miss!

After correcting that, I started generating the scenes one by one. For example, for the scene where I turn to the camera, the prompt (based on the AI's storyboard) was:

I spin my ergonomic chair around to face the camera. I push my glasses, which I don't have up to the bridge of my nose, and I say, 'I'm Claire and this is how I AI.'

For each prompt, the model generated two video options, which is a nice feature for getting some variety. The results were fascinating. Some were genuinely impressive, while others were... spooky. And the audio it generated for some of them was completely unexpected, like a robotic voice saying, "We were told AI would replace us."

Step 3: Assembling the Final Cut

Once all seven scenes were generated, the final step was to stitch them together. I didn't need to leave the browser or use a separate tool. Google Flow has a built-in video editor with a simple timeline.

I simply added my favorite version of each of the seven scenes to the timeline in the order the AI had suggested in the storyboard. I did some minor trimming and arrangement, and the whole editing process took maybe five minutes.

The Grand Premiere: The 'How I AI' Hype Video

And now, the moment of truth. After about 15 minutes of total work—from avatar creation to final edit—I had a complete, one-minute hype video.

So, how did it turn out? I'm actually obsessed with it. It’s not perfect, but it’s so much better than I expected for basically zero effort and zero video knowledge.

What Worked:

• Speed and Accessibility: This is the big one. To go from idea to a full video in minutes is wild. It truly democratizes video creation.
• Facial Accuracy (Sometimes): I'd say it was my face about 50% of the time. The side profiles were shockingly accurate. In one frame, it even captured the sun damage on my skin, so it’s not just smoothing everything out.

What Didn't Work (The Uncanny Valley):

• Character Inconsistency: My AI avatar had beautiful, long wavy hair, which I recently cut. The background also changed between scenes—sometimes books, sometimes plants, different wall colors. It pulled some real elements, like a poster from my real office, but couldn't keep the environment consistent.
• Emotional Expression: This is where it gets weird. A scene where I was supposed to be laughing was 100% uncanny valley. The expression was just… off. AI hasn't quite mastered authentic human emotion.

• Cheesy Tropes: The AI's idea of 'tech' is very early 2000s. I'm holding a 24-inch iPad and looking at a schematic of what looks like a church. It’s hilarious.
• Typography and Graphics: The text overlays and end cards were pretty lame. The design elements are definitely a weak point compared to the realism of the video generation.

Final Thoughts

Even with the quirks, I'm blown away. Is the technology 100% there? No. But is it 50% there and already incredibly useful? Absolutely. With a bit more effort—more descriptive prompting for background consistency, maybe providing more source images—I bet I could create a video that would convince most people it was real.

If this experiment proves anything, it's that we're entering an era where anyone can be a creator. I went from having zero ability to produce a video to creating a pretty compelling one in less time than it takes to drink a cup of coffee. If that’s not a How I AI success story, I don’t know what is.

I hope you enjoyed this strange mini-episode. I'm definitely going to be playing with the Google Omni model and Google Flow more. I would love to hear if you try it out, and please share your own AI-generated video examples in the comments!

---

Thanks to Our Sponsors

This episode of How I AI was brought to you by:

• Merge—Connective infrastructure for production AI
• Jira Product Discovery—Prioritize with insights, build with confidence

Episode Links

• Find me, Claire Vo, on LinkedIn and X.
• Try my free AI tool for product managers, ChatPRD.

Tools Referenced:

• Google Flow
• Gemini Omni
• Veo 3

Production and marketing by [Penname](https://penname.co/). For inquiries about sponsoring the podcast, email jordan@penname.co.