Week 27: Using AI to make a cartoon sourdough tutorial

The Experiment

After creating a video of myself hanging out with Jeff Bezos last October, I wanted to take my AI video skills to a different level. This time, my goal was to create a Simpsons-style animated tutorial about my true passion: sourdough bread making. The challenge was to create a short, shareable tutorial that would be fun, slightly weird, and test how much these tools have improved in the past six months.

The Process

 Here's how I went about it:

1. Brain Dump with Otter.ai

Started by speaking my thoughts into Otter to create a rough script This gave me raw material to work with and structure my ideas

2. Planning with ChatGPT

Asked ChatGPT to create a shot-by-shot overview Had it outline beginning and end frames for each section Got it to summarize everything into a neat storyboard

3. Generating Simpsons-Style Imagery

Used ChatGPT to create Simpsons-style illustrations of the bread-making process Requested beginning and end frames for each scene in the sourdough process Hit a snag when ChatGPT was down for a couple of hours (classic AI reliability)

4.Animation with Kling.ai

Fed the illustrations into Kling.ai to animate transitions between frames Had it create smooth movements between the beginning and end frames Discovered some fun glitches (like my hand remaining in the frame while my body disappears, see above

5. Assembling in iMovie

Imported all the animated sequences. Adjusted timing for each segment. Worked out duration for each section

6.Scripting with Claude

Had Claude create a final narration script based on my timing Some timing was a bit off, leaving a few "pregnant pauses" But overall, it created a workable narration

The Outcome

The final product: a minute-and-a-half animated tutorial on sourdough bread making, complete with Simpsons-style visuals and some charmingly glitchy transitions. While not perfect, it was surprisingly effective for a $7 investment and a few hours of work.

I could have iterated more and cleaned up the glitches, but I actually liked the slightly freaky, imperfect results. There's something about those AI imperfections that adds character to the final product (and I want to see how these change over time, whether AI cleans up the imperfections)

Captions are auto generated

Key Takeaway

AI video creation tools have definitely improved in the last six months, but they still have quirks that require workarounds. Kling.ai has gotten less glitchy overall, but you need to be strategic about your image inputs. The closer your beginning and end frames are in composition and positioning, the better the animation works between them.

Pro Tips for Beginners:

1. Frame Similarity Matters: Design your beginning and end frames with similar positioning of key elements. The less Kling has to "imagine" between frames, the cleaner the animation.

2. Start with a Solid Storyboard: Get ChatGPT to help map out your scenes before generating any images. This saves you from creating visuals that won't work well together.

3. Keep Animation Sequences Short: Aim for 5-10 second clips between keyframes. Longer sequences tend to introduce more glitches and oddities.

4. Have Backup Tools Ready: When one AI service goes down (like when ChatGPT wouldn't generate images), have alternatives ready to keep your momentum.

5. Don't Over-Polish: Sometimes the weird AI artifacts give your video character. A perfectly clean animation might actually feel less interesting than one with a few quirks.

Want to Try It Yourself?

• Otter.ai for transcribing your initial ideas

• ChatGPT for storyboarding and image generation

• Kling.ai for animation ($7 for basic use)

• iMovie or any basic video editor for final assembly

• Claude for script refinement

What's Next?

I plan to revisit this experiment in another six months to see how these tools continue to evolve. Will Kling.ai fix those transition glitches? Will image generation become more reliable? While I wait, I'll keep baking with my 1840 Oregon Trail starter (bread nerds hit me up to talk hydration levels etc, or I'm happy to share starter or tips for newcomers).

I'm also going to play with veo3 just to see what all the hubbub is all about, if you're lucky you'll see more ultra amateur video content coming your way.