by Tencent
Large-scale video generation model from Tencent. 13B parameter 3D DiT with Hunyuan-Large MLLM text encoder (~7B, not T5-based). Strong motion quality and visual fidelity up to 720p.
Your hardware
Detecting...
Measured quality metrics for HunyuanVideo outputs.
How often humans prefer this model's output (0-100%)
Visual quality and composition rating (5-9 scale)
VRAM estimates at FP16 and FP8 precision. FP8 uses ~40% less memory with minimal quality loss. Grade shows how well each GPU handles the generation workload.
| Scenario | VRAM | RTX 4090 24GB | RTX 3060 12GB | RTX 4060 8GB | MacBook Pro M4 Pro 24GB |
|---|---|---|---|---|---|
| 512×512 · 25 frames | 20.9 GB | A● | F● | F● | F● |
| 768×512 · 25 frames | 21.9 GB | A● | F● | F● | F● |
| 768×512 · 100 frames | 24.7 GB | B● | F● | F● | F● |
| 1280×720 · 25 frames | 25.7 GB | B● | F● | F● | F● |
| Scenario | VRAM | RTX 4090 24GB | RTX 3060 12GB | RTX 4060 8GB | MacBook Pro M4 Pro 24GB |
|---|---|---|---|---|---|
| 512×512 · 25 frames | 34.8 GB | F● | F● | F● | F● |
| 768×512 · 25 frames | 37.7 GB | F● | F● | F● | F● |
| 768×512 · 100 frames | 46.1 GB | F● | F● | F● | F● |
| 1280×720 · 25 frames | 49.0 GB | F● | F● | F● | F● |
Turbo / LCM distillation
Use distilled scheduler at 4-8 steps for faster iteration
from diffusers import HunyuanVideoPipeline
import torch
pipe = HunyuanVideoPipeline.from_pretrained(
"tencent/HunyuanVideo",
torch_dtype=torch.float16
)
pipe.to("cuda")
frames = pipe(
prompt="your prompt here",
num_inference_steps=50,
guidance_scale=6.0,
num_frames=129,
).frames[0]
# Save frames or export as videoGet started
Setup instructions for running HunyuanVideo locally
1. Download the model
Get the checkpoint from HuggingFace
2. Place in:
ComfyUI/models/checkpoints/3. Launch ComfyUI
python main.pyVRAM allocation for 25 frames at 768×512 on RTX 4090 24GB
25 frames at 768×512, 30 steps, FP16.
Download HunyuanVideo in different precisions. Lower precision = less VRAM but slight quality loss.
| Format | Precision | Size | Provider | |
|---|---|---|---|---|
| safetensorsRecommended | FP16 | 26.0 GB | official | Download |
Growing LoRA ecosystem with character and style LoRAs.
Frequently asked questions
HunyuanVideo (13B parameters) requires approximately 21.9 GB of VRAM at FP16 precision for generating 25 frames at 768×512. Video generation typically requires more VRAM than image generation due to temporal attention layers.
Yes, the RTX 4090 (24 GB VRAM) can run HunyuanVideo at FP16. Expected generation time is around ~3m 3s for a 25-frame clip.
On a reference GPU (RTX 4090 24GB), HunyuanVideo generates a 25-frame video at 768×512 in approximately ~3m 3s at FP16 with 30 inference steps. Faster GPUs with higher memory bandwidth will reduce generation time.
HunyuanVideo supports up to 1280×720 resolution and 129 frames per generation at 24 FPS. Higher resolutions and frame counts require proportionally more VRAM.
See also