whatmodel.sh / video / Hotshot-XL

Hotshot-XL

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

Hotshot-XL is an open-weights video model from Hotshot, released October 2023. Its published size is 2.8B parameters. Licence: openrail++.

Model information sources

· Parameter count ↗

UNet including temporal layers: HF Hub safetensors total for the official repo (pinned revision), 2,804,780,484 = unet/diffusion_pytorch_model.safetensors (the SDXL UNet's 2,567,463,684 plus the separately shipped 237,316,800-parameter temporal layers). The two CLIP text encoders and the VAE are not included. Observed 2026-10-07.

27.0±1.9#86 of 93 video modelsfrom 6 benchmark results across 2 sources

$ how it compares

The video models ranked either side of Hotshot-XL, out of 139 with a score.

#modelscorewhereRAMreleased
125NOVA-0.6B T2V BAAI29±3local~9.3GB2024-12
126ZeroScope v2 576w cerspense27±3local~7.9GB2023-06
127Hotshot-XL Hotshot27±2local~10GB2023-10
128Lumiere (T2V) Google27±4comparison only2024-01
129DynamiCrafter (1024) CUHK / Tencent AI Lab (DynamiCrafter)27±4comparison only2024-02

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

$ what it's good at

text to video▆▄▃▄▆▅█▃▂▂▂▄ #60 of 68 · 98% of its score · 6 tests

Beats ZeroScope v2 576w here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
AIGVE-60K human perception MOS (LOVE)42.7% ✎▄▄·█··█▄▄▄▄▄ beats 3/10easy · avg 73%▇▇▇▇▇▇▇▇ 20%at its level
EvalCrafter 2023 motion quality52.98 / 100 ✎▄····▄██··▄▄ beats 2/7very easy · avg 85.14 / 100▇▇▇▇▇▇▇ 18%at its level
EvalCrafter 2023 visual quality60.38 / 100 ✎▄█·····▄·█▄▄ beats 4/7very easy · avg 87.87 / 100▇▇▇▇▇▇▇ 17%at its level
EvalCrafter 2023 temporal consistency59.96 / 100 ✎▄···▄▄▄▄▄··█ beats 4/7very easy · avg 88.67 / 100▇▇▇▇▇▇▇ 17%at its level
EvalCrafter 2023 text-video alignment61.52 / 100 ✎▂··▂▂·····▂█ beats 5/7very easy · avg 87.86 / 100▇▇▇▇▇▇ 15%at its level
AIGVE-60K human text-video correspondence MOS (LOVE)47.8% ✎▄▄·█▄█▄··▄▄▄ beats 3/10easy · avg 70%▇▇▇▇▇ 12%at its level

$ where to run it

Other ways to run it

  1. Run it as published · needs ~10 GB, NVIDIA only · from github.com/hotshotco
    git clone https://github.com/hotshotco/Hotshot-XL && cd Hotshot-XL && pip install -r requirements.txt && python inference.py --prompt="a bulldog in the captains chair of a spaceship, hd, high quality" --output="output.gif"

Weights and documentation: huggingface.co/hotshotco ↗

$ how the score was worked out · EvalCrafter: original 2023 te… 67%, LOVE: Benchmarking and Evalua… 31%

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
EvalCrafter: original 2023 text-to-video evaluations4▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 67%CC-BY-4.0
LOVE: Benchmarking and Evaluating Text-to-Video Generation and Video-to-Text Interpretation (AIGVE-60K)2▇▇▇▇▇▇▇▇▇ 31%CC-BY-4.0
prior · pull toward the middle▇ 2%
weighted average= 27.0

$ 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