by BestWishYsh
14B distilled video model achieving 19.5 FPS real-time generation. Based on Wan2.1-T2V-14B with pyramid distillation for minute-scale coherent video. Apache 2.0 licensed.
Your hardware
Detecting...
Measured quality metrics for Helios 14B 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 | 44.2 GB | F | F | F | F |
| 768×512 · 25 frames | 46.3 GB | F | F | F | F |
| 768×512 · 100 frames | 52.6 GB | F | F | F | F |
| 1280×720 · 25 frames | 54.8 GB | F | F | F | F |
| Scenario | VRAM | RTX 4090 24GB | RTX 3060 12GB | RTX 4060 8GB | MacBook Pro M4 Pro 24GB |
|---|---|---|---|---|---|
| 512×512 · 25 frames | 24.6 GB | B | F | F | F |
| 768×512 · 25 frames | 26.7 GB | B | F | F | F |
| 768×512 · 100 frames | 33.0 GB | D | F | F | F |
| 1280×720 · 25 frames | 35.1 GB | F | F | F | F |
Turbo / LCM distillation
Use distilled scheduler at 4-8 steps for faster iteration
from diffusers import WanPipeline
import torch
pipe = WanPipeline.from_pretrained(
"BestWishYsh/Helios-Distilled",
torch_dtype=torch.float16
)
pipe.to("cuda")
frames = pipe(
prompt="your prompt here",
num_inference_steps=20,
guidance_scale=5.0,
num_frames=240,
).frames[0]
# Save frames or export as videoGet started
Setup instructions for running Helios 14B 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 Helios 14B in different precisions. Lower precision = less VRAM but slight quality loss.
| Format | Precision | Size | Provider | |
|---|---|---|---|---|
| safetensorsRecommended | BF16 | 28.0 GB | official | Download |
Compatible with some Wan 2.1 LoRAs.
Frequently asked questions
Helios 14B (14B parameters) requires approximately 46.3 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.
Helios 14B exceeds the RTX 4090's 24 GB VRAM at FP16 for video generation. Consider reducing resolution, frame count, or using a GPU with more VRAM.
On a reference GPU (RTX 4090 24GB), Helios 14B generates a 25-frame video at 768×512 in approximately ~3m 10s at FP16 with 30 inference steps. Faster GPUs with higher memory bandwidth will reduce generation time.
Helios 14B supports up to 1280×720 resolution and 240 frames per generation at 24 FPS. Higher resolutions and frame counts require proportionally more VRAM.
See also