QCaption: Video Captioning and Q&A through Fusion of Large Multimodal Models
Jiale Wang 1 , Gee Wah Ng 1 , Lee Onn Mak 1 , Randall Cher 2 , Ng Ding Hei Ryan 2 , Davis Wang 2 · 2026
Read QCaption: Video Captioning and Q&A through Fusion of Large Multimodal Models by Jiale Wang 1 , Gee Wah Ng 1 , Lee Onn Mak 1 , Randall Cher 2 , Ng Ding Hei Ryan 2 , Davis Wang 2 (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-06566.txt.
Description
Film is hard for AI to sum up, since a tale lives in both sight and sound over time. Jiale Wang and 5 co-authors built QCaption to do that job in a 2026 paper. It does not lean on one big model. It fuses three parts that each do what they do best.
First, a frame step picks the key stills from the film. A big sight-and-word model then reads those stills with the text. A big word model works the text once more to shape the end tale and to field asks on it. The whole set runs on site, with no cloud call, a plus for shops with strict rules. On tests, tale grade rose by a wide share, and ask grade rose by more. Part-by-part tests showed what the word model adds to the fuse.
Catalog record
| CREATOR | Jiale Wang 1 , Gee Wah Ng 1 , Lee Onn Mak 1 , Randall Cher 2 , Ng Ding Hei Ryan 2 , Davis Wang 2 |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 29 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-06566