Stable diffusion dreams of psychedelic faces

Andrej Karpathy · Intermediate ·🎨 Image & Video AI ·3y ago

Key Takeaways

Stable diffusion generates psychedelic faces from noise vectors using spherical interpolation, rendered on an A100 GPU.

Full Transcript

[Music] [Music] [Music] so [Music] [Music] [Music] [Music] [Music] [Music] do [Music] so [Music] [Music] you

Original Description

Prompt: "psychedelic faces" Stable diffusion takes a noise vector as input and samples an image. To create this video I smoothly (spherically) interpolate between randomly chosen noise vectors and render frames along the way. This video was produced by one A100 GPU taking about 10 tabs and dreaming about the prompt overnight (~8 hours). While I slept and dreamt about other things. Music: Stars by JVNA Links: - Stable diffusion: https://stability.ai/blog - Code used to make this video: https://gist.github.com/karpathy/00103b0037c5aaea32fe1da1af553355 - My twitter: https://twitter.com/karpathy
Watch on YouTube ↗ (saves to browser)
Sign in to unlock AI tutor explanation · ⚡30

Playlist

Uploads from Andrej Karpathy · Andrej Karpathy · 6 of 17

← Previous Next →

This video demonstrates how to use stable diffusion to generate psychedelic faces from noise vectors, showcasing the capabilities of AI in image generation. By smoothly interpolating between noise vectors, the model produces a dreamlike sequence of faces. The video highlights the potential of stable diffusion in creating unique and fascinating AI art.

Key Takeaways
  1. Choose a prompt, such as 'psychedelic faces'
  2. Generate random noise vectors
  3. Use spherical interpolation to smoothly transition between noise vectors
  4. Render frames using an A100 GPU
  5. Combine frames to create a video
💡 Spherical interpolation can be used to create smooth transitions between noise vectors, resulting in a dreamlike sequence of images.

Related Reads

📰
Teaching Kiro to Paint: A Stateful Image-Editing Skill Built on Gemini's Interactions API and MCP
Build a stateful image-editing skill using Gemini's Interactions API and MCP to generate and edit images with Kiro
Dev.to AI
 📰
I Tried to Replace My Image Workflow with an AI Photo Generator. Three Times It Failed. Once It Didn't.
Learn from an experiment replacing an image workflow with an AI photo generator, and discover the challenges and one success story
Dev.to AI
 📰
The Best Free AI Image Generators Better Than ChatGPT and Gemini
Discover free AI image generators that outperform ChatGPT and Gemini for specific workflows, offering superior photorealism and graphic design capabilities
Dev.to AI
 📰
50+ Sequential Images, One Prompt in Codex
Learn to generate sequential images with Codex using a single prompt and understand the limitations of this approach
Medium · ChatGPT
Up next
Logo Animations with AI 🤩 your Brand / Shop Promotions with AI 🚀 AI Videos Free #ai #aivideo
Raj Photo Editing and Much More
Watch →