2026-10-29

SenseTime open-sources Code2Video, a model that makes videos by writing code: worth a look for explainer and motion-graphics teams, not for live-action short dramas (late October)

SenseTime's SenseNova team published SenseNova 6.8 Code2Video on Hugging Face on October 26 (Beijing time): a 27-billion-parameter open model with a 512K context under Apache 2.0 that doesn't generate pixels but writes, renders and revises the program that draws each frame. Creators of explainers, title and caption packages, motion graphics or stylised 2D/3D shorts can put it on their trial list for work where text and timing must be exactly editable; teams making live-action or photorealistic AI short dramas can skip it for now. SenseTime shows curated examples only, with no benchmark, its deployment commands assume four GPUs, a single film takes hours of agent time, and we have not tested the output.