Text-to-image generation explained @GoogleResearch
Text-to-image generation explained  @GoogleResearch
Uploaded January 2023 | Updated September 2026, 2 weeks ago
Welcome to Hidden Layers - a series where we’ll show you how advanced ML algorithms from Google Research work in a way that’s easy to understand and accessible. In this episode, we get a behind-the-scenes look at how text-to-image models like Imagen and Parti operate. Watch this video to learn about the science behind these models and more!


Resources:

Check out Imagen → https://goo.gle/3yvFNsS
Check out Parti → https://goo.gle/3Vk5wi2
Try out the AI Test Kitchen → https://goo.gle/3XKsQG6

Chapters:
0:00 - Intro
0:40 - What is the science behind text-to-image models
2:23 - What is diffusion?
2:40 - What are the different architectures and approaches?
3:00 - What is an auto-regressive approach?
4:46 - What is the AI Test Kitchen?


Watch more episodes of Hidden Layers→ https://goo.gle/HiddenLayers
Subscribe to the Google Research Channel → https://goo.gle/GoogleResearch
Text-to-image generation explainedMapping the Web of Life | Google’s AI for BiodiversityQuantum Echoes: Verifiable Quantum Advantage | Behind the BreakthroughsWhat drives Google Research?What Happens When You Teach a Robot to Understand LanguageAfrican Languages Series: Get to know Yoruba, one of the main languages of NigeriaWho uses JAX?Google Research Vs. Inequity in Media | How AI is an Ally for RepresentationAfrican Languages Series: All about Akan, the main native language of the Akan people of GhanaHow LLMs might help scale world class healthcare to everyoneGrounding language in robotic affordancesWhy Robots Struggle to Pick Up Small Objects
Google Research |

Text-to-image generation explained

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER