May 20, 2026
Lance? More like comments at dawn
Show HN: Lance – image/video generation and understanding in one model
ByteDance says one AI does it all, but commenters are fighting over the name and the blurry videos
TLDR: ByteDance unveiled Lance, an AI that can both analyze and create images and videos in one system. Commenters instantly turned it into a drama thread, arguing over the confusing name, mocking the soft-looking video demos, and splitting between hype and hard-eye-roll skepticism.
ByteDance rolled into Hacker News with Lance, a new artificial intelligence model that claims it can both understand and create images and videos in one place. In plain English: it can answer questions about pictures and clips, generate new visuals from text, and even edit them. The company is pitching it as a lightweight all-in-one system, and one fan immediately chimed in with the classic open-source battle cry: “Great quality, forked and going to try.”
But the real show was in the comments, where the mood swung from impressed to brutally unimpressed in seconds. One of the first mini-dramas? The name. A commenter groaned that “Lance” is already tied to the popular LanceDB project, while another went full meme mode with “last dance for lance vance!” Suddenly the launch had accidental identity-crisis energy.
Then came the real roast: the video quality. Critics were not gentle. One commenter sneered that with massive resources behind it, the end result was basically “silly little slop videos,” which is about as polite as throwing tomatoes in internet form. Another called out the demos for looking low-resolution and smoothed over, asking why anyone would still be showing souped-up, not-quite-sharp video in 2026. So yes, ByteDance unveiled a flashy all-in-one image-and-video AI, but the crowd immediately split into two camps: “cool, I’m trying this now” and “why does this still look blurry?”
Key Points
- •ByteDance introduced Lance as a unified multimodal model for image and video understanding, generation, and editing.
- •The project says Lance uses only 3B active parameters while delivering strong results on image generation, image editing, and video generation benchmarks.
- •According to the article, Lance was trained entirely from scratch using a staged multi-task training recipe.
- •The article states that training was completed within a 128-A100-GPU compute budget.
- •Demo examples show capabilities across text-to-video, video editing, multi-turn consistency editing, video understanding, text-to-image generation, image editing, and image understanding.