Local video generation models
Video models you can run on your own GPU, with memory needs and generation times.
| Model | Made by | Backbone | Native output | Quality tier | Released |
|---|---|---|---|---|---|
| LTX-2.3 | Lightricks | 18.9B | 1280×704 · 121 frames | 86 | 2026-03 |
| Wan 2.2 I2V A14B | Alibaba | 14.0B | 1280×720 · 81 frames | 86 | 2025-07 |
| Wan 2.2 T2V A14B | Alibaba | 14.0B | 1280×720 · 81 frames | 86 | 2025-07 |
| HunyuanVideo 1.5 (8.3B) | Tencent | 8.3B | 1280×720 · 121 frames | 84 | 2025-11 |
| LTX-2 (19B, video + audio) | Lightricks | 18.9B | 1280×704 · 121 frames | 82 | 2026-01 |
| SkyReels V2 T2V 14B 720P | Skywork | 14.0B | 1280×720 · 97 frames | 78 | 2025-04 |
| Wan 2.1 I2V 14B 720P | Alibaba | 14.0B | 1280×720 · 81 frames | 78 | 2025-02 |
| Wan 2.1 T2V 14B | Alibaba | 14.0B | 1280×720 · 81 frames | 78 | 2025-02 |
| Wan 2.1 VACE 14B | Alibaba | 17.3B | 1280×720 · 81 frames | 78 | 2025-05 |
| Wan 2.2 Animate 14B | Alibaba | 17.3B | 1280×720 · 77 frames | 78 | 2025-09 |
| Step-Video T2V (30B) | StepFun | 30.0B | 992×544 · 204 frames | 76 | 2025-02 |
| Wan 2.2 S2V 14B (speech to video) | Alibaba | 16.3B | 832×480 · 81 frames | 76 | 2025-08 |
| FramePack I2V (HunyuanVideo 13B) | lllyasviel | 12.9B | 640×640 · 150 frames | 74 | 2025-04 |
| SkyReels V2 T2V 14B 540P | Skywork | 14.0B | 960×544 · 97 frames | 74 | 2025-04 |
| Wan 2.1 I2V 14B 480P | Alibaba | 14.0B | 832×480 · 81 frames | 74 | 2025-02 |
| HunyuanVideo 13B | Tencent | 13.0B | 1280×720 · 129 frames | 72 | 2024-12 |
| HunyuanVideo I2V 13B | Tencent | 13.0B | 1280×720 · 129 frames | 72 | 2025-03 |
| Cosmos-Predict2 14B Video2World | NVIDIA | 14.0B | 1280×704 · 93 frames | 70 | 2025-06 |
| Open-Sora 2.0 (11B) | HPC-AI Tech | 11.0B | 768×768 · 129 frames | 70 | 2025-03 |
| Wan 2.2 TI2V 5B | Alibaba | 5.0B | 1280×704 · 121 frames | 70 | 2025-07 |
| LTX-Video 13B 0.9.8 (distilled) | Lightricks | 13.0B | 1216×704 · 121 frames | 68 | 2025-07 |
| Mochi 1 Preview (10B) | Genmo | 10.0B | 848×480 · 163 frames | 62 | 2024-10 |
| CogVideoX 1.5 5B | Zhipu AI | 5.6B | 1360×768 · 81 frames | 60 | 2024-11 |
| Cosmos 1.0 Diffusion 7B Text2World | NVIDIA | 7.0B | 1280×704 · 121 frames | 60 | 2025-01 |
| Cosmos-Predict2 2B Video2World | NVIDIA | 2.0B | 1280×704 · 93 frames | 58 | 2025-06 |
| CogVideoX 5B | Zhipu AI | 5.0B | 720×480 · 49 frames | 55 | 2024-08 |
| CogVideoX 5B I2V | Zhipu AI | 5.6B | 720×480 · 49 frames | 55 | 2024-09 |
| Wan 2.1 T2V 1.3B | Alibaba | 1.3B | 832×480 · 81 frames | 55 | 2025-02 |
| Wan 2.1 VACE 1.3B | Alibaba | 2.2B | 832×480 · 81 frames | 55 | 2025-05 |
| Allegro (2.8B) | Rhymes AI | 2.8B | 1280×720 · 88 frames | 52 | 2024-10 |
| SkyReels V2 Diffusion Forcing 1.3B | Skywork | 1.4B | 960×544 · 97 frames | 52 | 2025-04 |
| LTX-Video 2B | Lightricks | 1.9B | 768×512 · 121 frames | 50 | 2024-11 |
| CogVideoX 2B | Zhipu AI | 1.7B | 720×480 · 49 frames | 45 | 2024-08 |
| AnimateDiff v3 (SD 1.5) | Shanghai AI Lab | 1.3B | 512×512 · 16 frames | 35 | 2023-12 |
Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.