Revolutionizing Education: Children Learn GenAI Through ComfyUI, Z-Image, and WanAnimate
A pioneering classroom initiative in Hungary is teaching children generative AI using open-source tools like ComfyUI, Z-Image, and WanAnimate, blending creativity with technical literacy. Educators report unprecedented engagement, though challenges in image editing persist.

Revolutionizing Education: Children Learn GenAI Through ComfyUI, Z-Image, and WanAnimate
In an innovative classroom experiment in Budapest, Hungary, children as young as eight are learning to create AI-generated art and animations using open-source tools such as ComfyUI, Z-Image, and WanAnimate. Led by Dr. Sándor Kovács, a former AI researcher turned educator, the program represents a groundbreaking shift in pedagogy—moving beyond passive digital consumption to active creation with generative artificial intelligence. The initiative, documented in a detailed public report on drsandor.net/ai/school/, showcases student-generated images and a final animated video produced by the teacher, demonstrating both child creativity and instructional scaffolding.
According to Britannica’s definition of teaching as “educating, mentoring, and facilitating,” the core function of educators is to help students learn by imparting knowledge and guiding discovery. Dr. Kovács’s approach aligns precisely with this principle: rather than lecturing on AI theory, he created a hands-on, project-based curriculum where students iteratively designed prompts, edited outputs, and collaborated on storytelling through visuals. The use of ComfyUI—a modular, node-based interface for Stable Diffusion—allowed students to visualize the AI generation pipeline without needing to code, lowering the barrier to entry. Z-Image enabled intuitive image editing, while WanAnimate transformed static images into short animated sequences, giving children a tangible sense of agency over digital creation.
What sets this initiative apart is its focus on process over product. Students were encouraged to document their failures as much as their successes. One 10-year-old student, for instance, spent three days trying to generate a “flying cat with rainbow wings” after repeatedly encountering distorted limbs. Through peer feedback and iterative prompting, she eventually achieved her vision. Such experiences mirror findings in educational psychology, where Britannica notes that “effective teaching theories emphasize experiential learning, metacognition, and resilience in problem-solving.” The classroom became a laboratory for trial, error, and refinement—core tenets of scientific inquiry.
However, the project also revealed persistent technical limitations. As Dr. Kovács noted in his report, no Flux model or Qwen Image Edit tool adequately resolved fine-grained editing challenges—such as preserving facial symmetry or adjusting background elements without artifacting. These hurdles underscore a broader industry gap: while generative AI models have advanced rapidly, their usability for non-experts, especially children, remains inconsistent. The absence of reliable, child-friendly editing interfaces suggests that current tools are not yet optimized for pedagogical contexts.
Still, the educational outcomes are compelling. Teachers observed increased confidence in digital expression, heightened curiosity about technology’s inner workings, and improved collaboration skills. One parent reported her child, previously disengaged in school, now spending evenings researching AI ethics on YouTube. This aligns with Britannica’s broader view of teaching as a dynamic, multifaceted role that extends beyond content delivery to nurturing critical thinking and lifelong learning.
As generative AI becomes ubiquitous, the question is no longer whether children should be exposed to it—but how they should be guided through it. Dr. Kovács’s model offers a replicable blueprint: prioritize accessibility, encourage experimentation, and embed ethical reflection. While commercial platforms dominate the consumer AI space, this grassroots initiative proves that open-source tools, when thoughtfully curated, can empower the next generation not just to use AI—but to understand and shape it.
Experts in educational technology caution that scaling such programs requires teacher training, infrastructure support, and policy alignment. Yet, as one student wrote in her reflection: “I didn’t know computers could dream. Now I know I can help them.”


