Hairstyle Morphs
A four-step workflow that produces short morph videos to give anyone a new hairstyle
chatgptclaude-designnano-bananaklingimage-generationvideo-generation
Loading content...
What It Is
This is a fun one. Short morph videos that turn a photo of someone into a video of them with a new hairstyle, stitched together from a handful of AI tools that all launched around the same week. Mostly an excuse to play with Nano Banana, Kling, and Claude Design (for the animation at the top of this page). Next step is to combine these tools into a single workflow (in n8n, Make, Poppy, or a custom app) using the Gemini and Kling APIs.
What I Learned
- Building generative AI toys with family is its own kind of magic — Making morph videos with family and friends felt like exactly that. We'd fight over who got to go next, describe a wild new hairstyle, then watch ourselves transform two minutes later. I tried a platinum blonde mullet, my friend got a pink mohawk, and my daughter chopped her never-cut back-length hair to a short bob. Examples below!
- Generative AI changes the input from "pick one" to "describe anything" — Similar to advancements in Ethan Mollick's otter test, a year ago the closest we could get to this kind of magic was a Snapchat or Zoom filter, selected from a list of presets. Generative AI moves the ceiling to "anything you can describe," and the jump is delightful once you've seen someone or something you know totally transform before your eyes.
- Use each AI tool for what it's best at, then combine them — A video model like Veo can produce a hairstyle morph straight from a prompt, but I got dramatically better likeness (especially facial features) by generating the still After portrait in an image model first and feeding it to the video model as the final frame. Iterating on a still is also cheaper, faster, and easier.
- The pace of new releases hasn't dulled the joy of discovering what they can do — I handed Claude Design the four workflow steps and a rough Mermaid diagram, expecting a clean rudimentary explainer animation. It made a neat, well-timed walkthrough that was more creative than I'd asked for (the polaroid, simplified mock UIs for each tool, appropriate sound effects), all in one shot. There are things I'd improve if I iterated, but I left the one-shot output up to show how strong first tries can be now.
Built With
ChatGPT drafts the image prompts, Nano Banana generates the After portraits, Kling.ai Video 2.1 builds the morph from the before and after images, and Claude Design one-shotted the animated explainer at the top of this page.
Claude Design one-shotted the animated explainer at the top of this page in under five minutes from the brief prompt and follow-up questions seen here.