Skip to content
Video

Wan2.1

State-of-the-art open video generation model lineup

2 min readupdated

Wan2.1 is a comprehensive and powerful open video generation model lineup from Wan-Team that sets new standards in content synthesis. The repository features the T2V-1.3B (Text-to-Video) version, built on the Flow Matching architecture (a modification of diffusion models) and capable of running on consumer video cards. The model outperforms many existing open and commercial alternatives, supporting high-quality video generation from text descriptions. It will become an excellent tool for creative teams, video makers, and researchers with limited computational resources.

Key features

  • SOTA performance: Demonstrates better results than many open alternatives and commercial solutions.
  • Consumer GPU support: The T2V-1.3B version requires only 8.19 GB of VRAM and runs on an RTX 4090.
  • Text generation on video: The first video model that confidently generates text in Chinese and English.
  • Efficient Wan-VAE: The proprietary 3D variational autoencoder encodes and decodes 1080P video of any length without losing temporal information.

What you can do

  • Generate a 5-second video in 480P resolution from a text prompt using a Python script or Diffusers integration.
  • Connect prompt expansion via Dashscope API or local Qwen models for detailed scene development.
  • Run the Gradio interactive web interface locally to experiment with video generation.

Example

import torch
from diffusers import AutoencoderKLWan, WanPipeline
from diffusers.utils import export_to_video

model_id = "Wan-AI/Wan2.1-T2V-1.3B-Diffusers"
vae = AutoencoderKLWan.from_pretrained(model_id, subfolder="vae", torch_dtype=torch.float32)
pipe = WanPipeline.from_pretrained(model_id, vae=vae, torch_dtype=torch.bfloat16)
pipe.to("cuda")

output = pipe(
    prompt="A cat walks on the grass, realistic",
    height=480,
    width=832,
    num_frames=81,
    guidance_scale=5.0
).frames[0]
export_to_video(output, "output.mp4", fps=15)
Read moreDetails
VRAM requirements: from 8.19 GB for the 1.3B version (on an RTX 4090, generation takes about 4 minutes without quantization).
Architecture: Diffusion Transformer (DiT) based on the Flow Matching framework with a T5 text encoder.
Features: it is recommended to use 480P resolution for the 1.3B model due to result stability.
Examples
Discuss Telegram 0 19

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 5.3 GB · ready files9.1 GB
  • BF16weights 2.6 GB6.0 GB
  • FP8 / INT8recommendedweights 1.3 GB4.5 GB
  • 4-bitweights 0.7 GB3.9 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · F32 — from RTX 3080

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # SadikNawazAI/SADIK-VIDEO-AI
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download SadikNawazAI/SADIK-VIDEO-AI --local-dir ~/models/SADIK-VIDEO-AI
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to SADIK-VIDEO-AI
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • SageAttention 3v2.2.0from October 28, 2025

    even faster on RTX 50 / Blackwell

    FP4 attention for Blackwell. Faster than SageAttention 2, but the authors suggest v2 when accuracy matters.

    How to enable: Build from the sageattention3_blackwell folder of the repo; in ComfyUI via KJNodes.

    Earlier generations (2)
    • SageAttention 2++ — Quantised (INT8/FP8) attention with no visible quality loss. The main speed-up for image and video generation.
    • SageAttention 1 — First version: INT8 for Q·K. Still works, but 2++ is faster and more accurate.
  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (3)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
    • xFormers — The older way to speed up attention, works almost everywhere. Sage or Flash are faster where available.
  • cache-dit (DBCache)v1.5.6from November 20, 2026new release

    1.5–3× at nearly the same quality

    Reuses block outputs between diffusion steps. Supports most DiT models.

    How to enable: pip install cache-dit; one line cache_dit.enable_cache(pipe) for diffusers.

    Earlier generations (2)
    • MagCache — Skips steps by magnitude of change. More accurate than TeaCache, no calibration.
    • TeaCache — The first popular diffusion step cache. Works, newer methods are more accurate.
  • LightX2V step distill0.6.1from November 9, 2026new release

    5–10× fewer steps

    Distilled LoRAs and an engine: video in 4–8 steps instead of 30–50.

    How to enable: Load their step-distill LoRA, CFG 1.0, 4–8 steps.

  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of SADIK-VIDEO-AI (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

SadikNawazAI/SADIK-VIDEO-AI
Author
SadikNawazAI
Category
Video
Released
Posted

On NeyroLearn

Hugging Face

190

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments