Best text-to-video models to run locally on a 16GB graphics card
A leaderboard of the best text-to-video models you can run locally on a 16GB graphics card (with 32GB of system RAM). HunyuanVideo 1.5 is #1 right now, followed by Wan 2.2 A14B and Wan 2.1 14B. Models are ranked on human votes and video generation tests. Each model shows the download that fits and the command to run it.
Scores use reviewed text to video evaluations only. Compare scores within this task. Models without task measurements are excluded.
+ more
$ terminal
↗ share
detecting your OS…
Score vs memory to run
● open weights ◆ API only ─┘ best score for the memory · x: GB of RAM / VRAM for the smallest download (log)
| # | model | task score↓ | where↓ | speed↓ | sources |
|---|
| # | model | task score↓ | where↓ | speed↓ | sources |
|---|---|---|---|---|---|
| 36 |  HunyuanVideo 1.5 · fp16 720p T2V + Qwen2.5-VL q8_0 (sd.cpp) | 54±2 | local (slow) |   | |
| 42 |  Wan 2.2 A14B · gguf-q4_k_m ×2 experts + umt5 q8_0 (sd.cpp) | 52±2 | local (slow) | | |
| 44 | Wan 2.1 14B · bf16 | 50±2 | local (slow) |  | |
| 46 |  Kandinsky 5.0 T2V Lite · bf16 | 49±2* | local (slow) | | |
| 48 |  Cosmos-1.0-Diffusion-7B-Text2World · bf16 | 48±3 | local (slow) | | |
| 52 | Wan 2.2 TI2V 5B · bf16 | 47±4 | local | | |
| 53 |  CogVideoX-5B · bf16 | 47±2 | local |  | |
| 57 |  Pyramid Flow (SD3) · bf16 | 45±3* | local | | |
| 58 |  T2V-Turbo-v2 · fp32 | 45±7*† | local | | |
| 61 | Wan 2.1 T2V 1.3B · fp16 + umt5 q8_0 (sd.cpp) | 43±4 | local | | |
| 62 |  LTX-2.3 · gguf-ud-q4_k_m 22B dev + Gemma 3 12B q4 (sd.cpp) | 43±7* | local (slow) | | |
| 63 |  Mochi 1 · default | 42±2 | local | | |
| 65 | HunyuanVideo · bf16 | 41±3 | local (slow) | | |
| 68 | T2V-Turbo (VC2) · fp32 | 41±3† | local | | |
| 70 | CogVideoX-2B · fp16 | 39±2 | local | | |
| 71 | CogVideoX1.5-5B · bf16 + sequential CPU offload | 38±3 | local (slow) | | |
| 72 |  Allegro · bf16 | 38±2 | local | | |
| 73 |  VideoCrafter2 · fp32 | 37±2 | local | | |
| 76 |  LaVie (base T2V) · fp16 | 35±2 | local | | |
| 77 | LTX-Video 0.9 (2B) · bf16 | 35±3* | local | | |
| 79 |  Open-Sora 1.1 · bf16 | 34±6* | local | | |
| 81 |  NOVA-0.6B T2V · fp16 | 32±3* | local | | |
| 82 |  Hotshot-XL · fp16 | 31±2 | local | | |
| 85 | ZeroScope v2 576w · fp16 | 30±3 | local | | |
| 87 | VideoCrafter1 T2V · official script (checkpoint as stored) | 29±2 | local | | |
| 90 |  ModelScope T2V 1.7B · fp16 | 28±2 | local | | |
| 91 |  LVDM (text-to-video) · fp32 | 28±3* | local | | |
| 93 | ZeroScope v2 XL (576w → XL pipeline) · fp16 + model CPU offload (576w base, then XL vid2vid) | 27±2 | local | | |
| — | Kandinsky 6.0 Video Lite · bf16 + model CPU offload | not yet calibrated | local (slow) | | |
| — | LTX-2.5 · fp8 | not yet calibrated | local | |
+ 67 more don't fit your machine or budget · show them
97 of 97 catalogue models listed here · sources and licences
* new or thinly tested: ranked on too few benchmark results · † only the maker’s own numbers so far · ~RAM is estimated; full context may need more memory