whatmodel.sh / video / T2V-Turbo-v2

T2V-Turbo-v2

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

T2V-Turbo-v2 is an open-weights video model from UC Santa Barbara, released October 2024. Licence: apache-2.0.

37.9±4.9#72 of 93 video modelsfrom 2 benchmark results across 2 sources* thinly tested, treat the rank as provisional

$ how it compares

The video models ranked either side of T2V-Turbo-v2, out of 139 with a score, then more from UC Santa Barbara.

#modelscorewhereRAMreleased
105HunyuanVideo Tencent39±3api price unknown +local~28GB2024-12
106Pandora (world model) Maitrix.org (MBZUAI / UC San Diego)39±7comparison only2024-05
107T2V-Turbo-v2 UC Santa Barbara38±5api price unknown +local~9.4GB2024-10
108Vidu 1.5 ShengShu (Vidu)38±3comparison only2024-11
109T2V-Turbo (VC2) UC Santa Barbara38±3local~9.4GB2024-05
more from UC Santa Barbara
104T2V-Turbo (MS) UC Santa Barbara39±6comparison only2024-05

see all video models · find the ones that fit your machine

$ what it's good at

text to video▆▄▃▄▆▅█▃▂▂▂▄ #44 of 68 * · 48% of its score · 1 test

Beats Open-Sora 2.0, Wan 2.1 T2V 1.3B, Mochi 1 here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
VBench (T2V total) lab83.5% ✎█···▄▄▄····▄ beats 3/5very easy · avg 90%▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 48%small run
other tests1 not in a task · 42% of its score
testresultvs the others in its rundifficultycountswhy
WorldModelBench (human total)6.22 / 10 ✎▅··▅▃···▃··█ beats 1/8easy · avg 7.71 / 10▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 42%at its level

$ where to run it

Through an API

api · Replicate · price unknown · replicate.com ↗Price unknown. Run/API page linked from the authors' README; compute-time pricing is not converted to output-unit pricing.

Other ways to run it

version

  1. Run it as published · needs ~9.4 GB, NVIDIA only · from github.com/Ji4chenLi
    git clone https://github.com/Ji4chenLi/t2v-turbo && cd t2v-turbo && pip install gradio==3.48.0 && hf download VideoCrafter/VideoCrafter2 model.ckpt --local-dir vc2 && hf download jiachenli-ucsb/T2V-Turbo-v2 unet_mg.pt --local-dir t2v-turbo-v2 && python app.py --unet_dir t2v-turbo-v2/unet_mg.pt --base_model_dir vc2/model.ckpt --version v2 --motion_gs 0.0

Weights and documentation: huggingface.co/jiachenli-ucsb ↗

$ how the score was worked out · T2V-Turbo-v2: Enhancing Video… 48%, WorldModelBench: Judging Vide… 42%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
T2V-Turbo-v2: Enhancing Video Generation Model Post-Training through Data, Reward, and Conditional Guidance Design (arXiv 2410.05677v3) · lab-reported1▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 48%CC BY 4.0
WorldModelBench: Judging Video Generation Models As World Models1▇▇▇▇▇▇▇▇▇▇▇▇▇ 42%CC-BY-4.0
prior · pull toward the middle▇▇▇ 10%
weighted average= 37.9

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json