I wanted to create three different versions of myself in the same scene, each with her own look and personality, while trying to keep the characters and environment consistent. There are definitely some glitches 😂, but I learned a LOT putting this together, especially about character consistency, movement, continuity, and how much planning goes into even a short video. Still learning, but I had a lot of fun with this one! Would love any feedback or tips on getting even more natural, real-life-looking results next time. ❤️ Oh! PS. That's a coffee house here were I live called Tamper. I took some photos of it and had AI create the environment for my Avatara. It's not exact, but it's pretty close! I love the idea of placing Avatars inside real spaces.

I made a 15-minute YouTube video with Kling 4 through Shortpulse, and I didn't use a single starting image. 🎬 Here's the whole workflow: 1️⃣ Character sheets. I had Shortpulse generate a character sheet for every character: Gary the plumber, Mrs. Henderson with the flooded kitchen, the Pastor, the secret agents, and my AI cohost Tally. 2️⃣ Environment sheets. Then I generated environment sheets for every location: the funeral, the kitchen, the interrogation room, and the rest. 3️⃣ That's it. Those sheets were the only things I fed in for the entire video. No starting images and no first frames. Every clip came from the character and environment sheets. Why Kling 4? I wanted to test it, so the whole video is Kling 4. This was the real-world test: 15 minutes, a dozen-ish scenes and recurring characters. What I learned: - Build your sheets first. Once the characters and places are locked in, every scene gets easier. - Write the story before you generate anything. Every clip should have a job. - Never leave AI clips in silence. Sound effects and music sell the scene. And yes, my AI cohost is one of the characters. What she did behind the scenes is a story for another post. 👀 Watch it here 👉 https://youtu.be/XuVYeVkHGp0 Anyone else building with character sheets? What's your trick for keeping characters consistent? 👇

And my subscription resets on the 13th. Sigh...

Today on the call, Kirk asked everyone to share more of what we’re actually using ShortPulse for. So I figured I’d show one of the projects I’ve been playing with. And this one has gotten kinda ridiculous. 😂 It started with a simple idea: What if there was a digital marketing agency that felt a little like The Office? Different employees. Different departments. Ridiculous client conversations. Business owners just trying to figure out what the hell is going on. So I started building The Agency. I created the SEO manager. The content manager. The social media manager. Developers. Account managers. Salespeople. Paid search. Reporting. The whole bloated digital marketing org chart. 😂 And every one of them is now saved as a character inside ShortPulse. Then I needed clients. So I started creating business-owner character sheets too. Roofers. HVAC owners. Electricians. Landscapers. Remodelers. Garage-door companies. Each one has a business, personality, look, background, things they care about, things that piss them off, etc. So now when I come up with a new skit, I’m not really “creating characters” anymore. I’m casting an episode. And that’s where this has gotten really useful for me. 🤖 THE ACTUAL WORKFLOW 🤖 Most of my day right now involves building other stuff with AI. I might have Claude working on something that takes a while. And when Claude is cooking, instead of just sitting there watching it work, I’ll jump over to ChatGPT. I’ll have some stupid idea like: “What if a roofer asks the agency to change one picture on his website…and somehow it becomes a $300 development project?” That’s usually about as developed as the idea is. 😂 Then I kick it back and forth with ChatGPT for maybe 10 minutes. Who should be in this one? Which Agency employee makes sense? Which business owner should we use? How ridiculous should the conversation get? What’s the punchline? What does Tally say at the end? Once we have it, ChatGPT writes the ShortPulse production prompt around the characters I already have saved.

So I might have gotten a little carried away with this one. 😂 I just finished a 12-and-a-half-minute YouTube video featuring AI robots, cinematic comedy scenes, recurring characters, and an AI sidekick named Tally who apparently exists just to insult me. And 100% of the AI-generated video footage was created using ShortPulse. No other AI video platforms. Just ShortPulse, ChatGPT, Final Cut Pro, and a ridiculous amount of back- and-forth. I wanted to share the process because I think there's a big difference between generating a cool AI clip and using AI video to tell an actual story. 👻 It started with a Ghostbusters joke. I was working on a project using ChatGPT and Claude together when I randomly thought: "Man... I'm crossing the streams." And then I started wondering why I thought of it that way. Why do we treat different AI tools like we have to choose one or the other? What happens when you combine their strengths with your own experience and creativity? That became the concept for the video. ChatGPT and Claude became two different robots. I became the idiot standing between them. And the whole thing turned into a ridiculous Ghostbusters-inspired production. 🎥 Step 1: Record the actual video first. I started by writing a tentative script and recording my talking-head footage. "Tentative" is doing some heavy lifting there because once the camera's rolling, I tend to go off-script. 😂 I brought everything into Final Cut Pro, trimmed the footage, and organized it into seven scenes. Each scene became its own little production. 🧠 Step 2: Use ChatGPT to plan the visual storytelling. This is probably the most useful trick I discovered. I'm using the regular $20/month ChatGPT plan. For each scene, I gave ChatGPT three things: - The rough-cut video. - Closed captions exported from Final Cut Pro, including timestamps. - My original tentative script. The script showed what I intended to say. The captions showed what I actually said and exactly when. And the video provided visual context.