2.2M views · Jul 25, 2025 · 37:20 · 1 HN recommender · 3 mentions · discussed 2025–2026
A guest video by Welch Labs on 3Blue1Brown explaining how text-to-image and video models work. It covers CLIP, shared embedding spaces, diffusion models (DDPM, DDIM), conditioning, guidance and negative prompts.
Why Hacker News points people here: HN uses it as a clear explainer for diffusion mechanics, such as negative prompts and faster sampling, and for joint text-image encoders.
“WL & 3b1b made a nice video on it”HN comment, 2026
“WL & 3b1b made a nice video on it”
Nice video on training text and image encoders together.
> Because you somehow need a giant training set which describes images in natural language, no? That's definitely one way - they train a text encoder together with an image encoder on a labelled set of images. WL & 3b1b made a nice video on it: this video
“they literally subtract the prompt for 'low quality video' from the input, and it improves the quality.”
Shows subtracting the 'low quality video' prompt to improve generation.
My favourite instantiation of this weird invocation is from this AI video generator, where they literally subtract the prompt for 'low quality video' from the input, and it improves the quality. this video I've just migrated my AI product to a different underlying model and had to redo a few of the prompts that the new model was interpreting differently. It's not obseleted, just requires a bit of migration. The improved quality of the new models outweighs any issues around prompting.
“Breakthrough in image generation speed literally came from applying better differential equations for diffusion”
Explains diffusion speedups from statistical mechanics-based differential equations.
Breakthrough in image generation speed literally came from applying better differential equations for diffusion taken from statistical mechanics physics papers: this video
