NotebookLM Stickman Animation Engine
Video by The AI Garage — A complete walkthrough of building a free, unlimited stickman animation video pipeline using NotebookLM, ChatGPT, Google Flow, Google AI Studio, YouTube Audio Library, and CapCut.
Full Transcript
Starting a faceless YouTube channel sounds simple until you actually try. The scripting, the visuals, the voice over, the editing, it adds up fast. What takes a professional an hour can take a beginner an entire day. What if the whole pipeline ran on free tools and took a fraction of that time? Today, I will show you how to turn Notebook LM into a full stickman video engine.
From topic research all the way to a finished edited video. Stickman was the most requested niche in the comments, so here we are. No camera, no animation skills, no paid tools. And at the end, you will see the actual video this pipeline produces. Let’s build it.
Step one, topic and title. I open Notebook LM, create a new project, then head to YouTube. I have three stickman channels ready. Simple Ways of Life, Zen, and Productive Peter. I start with Simple Ways of Life. Copy the channel URL. Go back to Notebook LM and paste it as a source. Same for Zen. Same for Productive Peter. Three sources in. I hit next and let it load.
Now, this is where it gets interesting. I go to settings, chat, then custom. But before I paste the prompt, let me show you what it looks like. You can see it on screen now. This is the topic and title selector prompt. Grab it from the pinned comment if you want it. Once I save this, Notebook LM basically becomes a topic researcher for this exact niche. It knows what these channels talk about, how they write titles, what hooks their audience. I go to chat and type, “Give me five topic and title ideas for my next stickman video.” That is it. One line. Because the role prompt already did the setup work.
And here is something I like about this. Before it gives the titles, Notebook LM actually runs a quick analysis of the sources. It breaks down the themes, the content patterns, what keeps working across all three channels. I scroll through that first. Then the five ideas drop. Here they are. And look, none of these exist on those channels. Notebook LM read the patterns, the title structures, the angles that work, and built something new from them. That is why this is fully demonetization safe. You are not copying content. You are just building on what already works.
I am going with number two. How to rewire your brain to crave hard tasks. I looked at all five, but honestly, this one hit a little too close to home. I have been avoiding recording this video for two days.
Before step two, one quick thing. I put together a guide called the faceless YouTube engine. Nine stages, no fluff. Each stage has a checklist and homework, so you actually move forward. A sample video is built right alongside you throughout the guide, so nothing stays theoretical. If you have never made a video before, this is where you start. Link is in the description.
Step two, script time. I open a fresh Notebook LM project, head over to Simple Ways of Life, and filter the channel by popular. I want the top performing videos. These are the ones that proved the style works. I copy the top five URLs and paste them into Notebook LM one by one. Five sources in. Hit next, let it load. Then settings, chat, custom, same move as before. This time I paste in the script writer role prompt, which is in the description. Now Notebook LM has absorbed the storytelling style of those five videos. The pacing, the way information builds, how the hook opens. It is not going to copy them. It is going to write in that style with completely new content.
I go to chat and type, “Create a 1,000-word YouTube script about how to rewire your brain to crave hard tasks.” One line. That is all it needs because the role prompt already set everything up. And look at what comes back. This is not a generic AI script. Look at how it is structured. It opens with the introduction, the automation of laziness. Then it moves into the trap of cheap dopamine. Then the morning dopamine spike and crash. And it closes with practical neuroscience hacks to rewire your brain. Each section builds on the last. It reads like something a real creator wrote. That structure, that escalation, that came from studying what actually works on these channels.
And because we are copying the style, not the content, this script is completely original. Clean channel. No demonetization risk.
Step three, visual plan. I open ChatGPT and head to my prompts doc. I copy the visual planner prompt. This prompt takes your script and turns it into a complete visual plan with image prompts and motion prompts for every scene. And each scene is capped at 8 seconds maximum, since short scenes keep the pacing tight and the viewer locked in. I paste it into ChatGPT. The prompt has a placeholder for the script, so I go back to NotebookLM, where we generated the script in step two, copy it, and paste it into ChatGPT in place of the placeholder. I hit generate.
Look at what comes out. The first thing it produces is a visual base prompt. This is the consistency anchor. It defines the character style, the vibe, the background. Everything in the video will match this. Then it moves into the scenes. Each scene has three things: a voiceover, an image prompt, and a motion prompt. Every single scene follows that same structure, clean, ready to use.
Now, let me scroll down to scene five. This one has no stick man, and that is completely intentional. Good stick man videos do not use a stick man for every single scene. That gets repetitive fast. So, the Visual Planner prompt builds in roughly 20% of scenes as graphic visuals, stats, concepts, abstract ideas. It breaks the pattern just enough to keep the viewer watching.
Step four, visuals. I search for Google Flow on Google and open the first result. Since I am already logged in, it takes me straight to the dashboard. If this is your first time, just sign up with your Gmail account and you are in. I open a new project. Make sure image creation is selected, and Nano Banana is the model. Then, I go back to ChatGPT, copy the visual base prompt from the last step, paste it into Flow, and hit generate. Here is the result. The stick man comes out in different poses across the variations, and even though it is a simple character, this base generation matters. I will show you exactly why in a moment.
I upload that base generation, copy the image prompt for scene one from ChatGPT, paste it into Flow, and generate. And look, same stick man style, same lines, same feel. It picked up the character from the base and carried it into the scene. Then, I switch to video generation in Flow, upload the scene image, copy the motion prompt for scene one, paste it in, and generate. Here is the animation. This is why keeping motion prompts simple is important. Clean instructions produce clean movement. No noise, no weird artifacts, just a smooth stick man scene.
I do the same for scene two, image first, then animation. Look at the line thickness on this one. Same style as scene one. That consistency is a small detail, but it makes a real difference to the viewer’s eye. Even if they cannot explain why, they feel it. I continue the same process for remaining scenes.
One more thing before the final step. My guide, the faceless YouTube engine, maps this entire production process. From picking your niche all the way through publishing consistently. Every chapter ends with homework. And a real sample video is built with you at every stage. So, nothing stays abstract. If you are already producing videos and want to cut your production time significantly, this guide has the full system for that. Link is in the description.
Let’s continue with our stick man video creation. For voiceover, I open Google AI Studio and go to the speaker settings on the right panel. I click the drop-down and go through a few voice demos. Since this is a stick man story video, I want something warm and approachable. So, select a voice with a friendly tag. I copy the voiceover for the first few scenes from Notebook LM, where we generated the script. Paste it into Google AI Studio and hit run.
I download it and continue doing the same for the remaining parts of the script.
Now for background music. I go to YouTube Studio and open the audio library from the left menu. It is right there in the sidebar. Easy to miss if you have not used it before. Since this is a stick man story video, I want something light and positive. So, I filter by genre, ambient, and mood, bright. I scroll through the results and listen to a few. I find one that fits, download it, and move to the edit.
I open CapCut. This is where everything comes together. I drop the voice over in first. That is the skeleton of the edit. Then I upload all the scenes. Now I pull up the visual plan we generated before and use it as my map. It tells me exactly which scene goes where. I start syncing scenes to the voice over just by following the plan. Honestly, this part is very easy because all the thinking is already done.
Since the scenes run a bit longer than the voice over, I either trim or speed them up slightly to fit. And for transitions, I keep it to cold cuts, no effects. That hard cut style gives the video a slightly nostalgic storytelling feel that works really well for this format. Then I drop the music in and pull the volume down so the voice over stays in front.
One more thing worth mentioning, when Google Flow generates the video clips, it also generates sound effects automatically. That is actually a nice bonus. Small sounds tied to the movement make the scenes feel more alive and keep the viewer more engaged. After adding the music, I go through each scene and adjust the sound effect levels so nothing is clashing.
Here is something worth mentioning. I edited the first five scenes, roughly 30 seconds of video, in about 2 minutes. A full video at that pace would take around 30 minutes to edit. That is entirely because of the visual plan. Having that map makes the whole edit fast.
All right, let’s see the result.
That is the full pipeline. If anything in this process was unclear, drop your question in the comments. I go through every single one. And if you want to see more niche breakdowns like this, just comment the niche you want next. That is how I choose what to make. If you want to discover faceless content creation in another niche, click the video on your screen right now. It is one of the most viewed videos on the channel. See you there.
Pipeline Summary
| Stage | Tool | Purpose |
|---|---|---|
| Topic & Title | NotebookLM + YouTube | Research successful channels, generate topic & title ideas |
| Script | NotebookLM | Generate 1,000-word script in the style of top-performing videos |
| Visual Plan | ChatGPT | Break script into scenes with image + motion prompts |
| Character & Scenes | Google Flow (Nano Banana) | Generate consistent stickman character and scene images |
| Animation | Google Flow | Animate scenes from images + motion prompts |
| Voiceover | Google AI Studio | Generate voiceover from script |
| Music | YouTube Audio Library | Royalty-free background music |
| Editing | CapCut (free) | Sync scenes to voiceover, add music, export |