Catalog · Video generation
Best local video generation models
Local text-to-video and image-to-video models, ranked by peak VRAM. Video is the most memory-hungry modality, and most needs a discrete GPU.
- Models
- 11
- Lightest
- ~6 GB
- Heaviest
- ~20 GB
Models
- Wan 2.1 T2V 1.3B832×480 (480p) · 81 frames~6 GBQ4 GGUFRuns on: MacNVIDIAAMD · offload floor ~5 GB
Tools: ComfyUI, Diffusers
- CogVideoX-2B720×480 · 49 frames~8 GBfp16 + offloadRuns on: NVIDIAAMD · offload floor ~4 GB
Tools: Diffusers, ComfyUI
- Stable Video Diffusion (img2vid-XT)1024×576 · 25 frames~8 GBfp16 + offloadRuns on: MacNVIDIAAMD
Tools: ComfyUI, Diffusers
- Wan 2.2 TI2V 5B1280×704 (720p) · 121 frames~8 GBQ4 GGUFRuns on: MacNVIDIAAMD · offload floor ~5 GB
Tools: ComfyUI, Diffusers
- LTX-Video 2B1216×704 · 121 frames~10 GBfp8 + offloadRuns on: MacNVIDIAAMD · offload floor ~6 GB
Tools: ComfyUI, Diffusers
- Wan 2.1 T2V 14B1280×720 (720p) · 81 frames~12 GBQ4 GGUFRuns on: MacNVIDIAAMD · offload floor ~8 GB
Tools: ComfyUI, Diffusers
- CogVideoX-5B720×480 · 49 frames~16 GBINT8 / fp8Runs on: NVIDIAAMD · offload floor ~5 GB
Tools: Diffusers, ComfyUI
- HunyuanVideo544×960 · 129 frames~16 GBQ4 GGUFRuns on: NVIDIAAMD · offload floor ~8 GB
Tools: ComfyUI
- Wan 2.2 T2V A14B1280×720 (720p) · 81 frames~16 GBQ4 GGUFRuns on: NVIDIAAMD · offload floor ~8 GB
Tools: ComfyUI, Diffusers
- LTX-Video 13B1216×704 · 161 frames~20 GBfp8Runs on: MacNVIDIAAMD · offload floor ~12 GB
Tools: ComfyUI, Diffusers
- Mochi 1480×848 · 85 frames~20 GBfp8 + offloadRuns on: NVIDIA · offload floor ~18 GB
Tools: ComfyUI, Diffusers
Peak VRAM is the memory a run consumes, the same basis the site uses everywhere; see the methodology. To check a model against your exact device, open its compatibility page.
FAQ
What is the most memory-efficient local video generation model?
Wan 2.1 T2V 1.3B uses the least: about 6 GB at Q4 GGUF. With CPU offload it can drop to ~5 GB, more slowly.
How much GPU memory do I need for local video generation?
It ranges from about 6 GB to 20 GB of peak VRAM across the models here. The figure is the memory a run consumes, not the size of card you must buy, so match it to your usable VRAM with a gigabyte or two of margin.
Is the memory figure the download size or the run size?
The run size: peak VRAM actually consumed during generation, which is the number that decides if it fits. Diffusion models can also offload parts to system RAM to run on less, slower. Every figure here is sourced.