Wan14BT2VFusioniX / README.md

vrgamedevgirl84

Update README.md

a4fb9f8 verified 6 months ago

preview code

raw

history blame

9.17 kB

metadata

tags:
  - text-to-video
  - diffusion
  - merged-model
  - video-generation
  - wan2.1
widget:
  - text: >-
      Prompt: A gritty close-up of an elven princess kneeling in a rocky ravine,
      calming a wounded, desert dragon. Its scales are cracked, dry, She wears a
      crimson sash over bone-colored armor, her auburn hair half-tied back. The
      camera dollies in rapidly as she reaches for its eye ridge. Lighting comes
      from golden sunlight reflecting off surrounding rock, casting a warm,
      earthy hue with no artificial glow.
    output:
      url: videos/Video_00063.mp4
  - text: >-
      Prompt: Tight close-up of her smiling lips and sparkling eyes, catching
      golden hour sunlight. She wears a white sundress with floral prints and a
      wide-brimmed straw hat. Camera pulls back in a dolly motion, revealing her
      twirling under a cherry blossom tree. Petals flutter in the air, casting
      playful shadows. Soft lens flares enhance the euphoric, dreamlike vibe.
      (Before vs After — Left: Wan2.1 | Right: Merged model
      Wan14BT2V_MasterModel)
    output:
      url: videos/AnimateDiff_00001.mp4
  - text: >-
      Prompt: A gritty close-up of a dwarven beastmaster’s face, his grey beard
      braided tightly, brows furrowed as he looks just off-camera. The camera
      dollies out over his shoulder, revealing a perched gryphon watching him
      from a boulder, its feathers rustling slightly in the breeze. The moment
      holds stillness and mutual trust. Lighting is early daylight, clean and
      sharp with strong environmental clarity.
    output:
      url: videos/FusionX_00012.mp4
  - text: >-
      Prompt: A gritty close-up of a jungle tracker crouching low, face flushed
      with focus as she watches a perched macaw a few feet ahead. Her cheek
      twitches as she shifts forward, beads of sweat visible on her brow. The
      camera slowly dollies in from below her line of sight, capturing the
      moment her eyes widen in fascination. Lighting is rich and directional
      from above, creating a warm glow over her face with minimal shadows.
    output:
      url: videos/FusionX_00005.mp4
  - text: >-
      Prompt: A gritty close-up of a battle-worn ranger kneeling in a scorched
      clearing, calming a wounded gryphon whose wing is torn and bloodied. Its
      feathers are dusky bronze with streaks of ash-gray. She wears soot-covered
      hunter green armor, her blonde hair pulled into a loose braid. The camera
      dollies in as her hand brushes the creature's sharp beak. Lighting comes
      from late afternoon sun filtering through smoke, casting a burnt-orange
      haze across the frame.
    output:
      url: videos/Video_00069.mp4
base_model:
  - Wan-AI/Wan2.1-T2V-14B
license: apache-2.0

🌀 Wan2.1_14B_FusionX

High-Performance Merged Text-to-Video Model
Built on WAN 2.1 and fused with research-grade components for cinematic motion, detail, and speed — optimized for ComfyUI and rapid iteration in as few as 6 steps.

Merged models for faster, richer motion & detail — high performance even at just 8 steps.

📌 Important: To match the quality shown here, use the linked workflows or make sure to follow the recommended settings outlined below.

🌀 Preview Gallery

These are compressed GIF previews for quick viewing — final video outputs are higher quality.

📂 Workflows & Model Downloads

💡 ComfyUI workflows can be found here:
👉 Workflow Collection (WIP)
📦 Model files (T2V, I2V, Phantom, VACE):
👉 Main Hugging Face Repo

🧠 GGUF Variants:

🎬 Example Videos

Want to see what FusionX can do? Check out these real outputs generated using the latest workflows and settings:

Text-to-Video
👉 Watch Examples
Image-to-Video
👉 Watch Examples
Phantom Mode
👉 Watch Examples
VACE Integration
👉 Watch Examples

🚀 Overview

A powerful text-to-video model built on top of WAN 2.1 14B, merged with several research-grade models to boost:

Motion quality
Scene consistency
Visual detail

Comparable with closed-source solutions, but open and optimized for ComfyUI workflows.

💡 Inside the Fusion

This model includes the following merged components:

CausVid – Causal motion modeling for better flow and dynamics
AccVideo – Better temporal alignment and speed boost
MoviiGen1.1 – Cinematic smoothness and lighting
MPS Reward LoRA – Tuned for motion and detail
Custom LoRAs – For texture, clarity, and facial enhancements

All merged models use permissive open licenses (Apache 2.0 / MIT).

🔧 Usage Details

Text-to-Video

CGF: Must be set to 1
Shift:
- 1024x576: Start at 1
- 1080x720: Start at 2
- For realism → lower values
- For stylized → test 3–9
Scheduler:
- Recommended: uni_pc
- Alternative: flowmatch_causvid (better for some details)

Image-to-Video

CGF: 1
Shift: 2 works best in most cases
Scheduler:
- Recommended: dmp++_sde/beta
To boost motion and reduce slow-mo effect:
- Frame count: 121
- FPS: 24

🛠 Technical Notes

Works in as few as 6 steps
Best quality at 8–10 steps
Drop-in replacement for Wan2.1-T2V-14B
Up to 50% faster rendering, especially with SageAttn
Works natively and with Kaji Wan Wrapper
Wrapper GitHub
Do not re-add merged LoRAs (CausVid, AccVideo, MPS)
Feel free to add other LoRAs for style/variation
Native WAN workflows also supported (slightly slower)

🧪 Performance Tips

RTX 5090 → ~138 sec/video at 1024x576 / 81 frames
If VRAM is limited:
- Enable block swapping
- Start with 5 blocks and adjust as needed
Use SageAttn for ~30% speedup (wrapper only)
Do not use teacache
"Enhance a video" (tested): Adds vibrance (try values 2–4)
"SLG" not tested — feel free to explore

🧠 Prompt Help

Want better cinematic prompts? Try the WAN Cinematic Video Prompt Generator GPT — it adds visual richness and makes a big difference in quality. Download Here

📣 Join The Community

We’re building a friendly space to chat, share outputs, and get help.

Motion LoRAs coming soon
Tips, updates, and support from other users

👉 Join the Discord

⚖️ License

Merged under permissive licenses:

Apache 2.0 / MIT
You can use, modify, and redistribute
You must retain original license info
Outputs are not necessarily licensed — do your due diligence

This model is for research, education, and personal use only. Commercial use is your own responsibility. Please consult a legal advisor before monetizing outputs.

🙏 Credits

WAN Team (base model)
aejion (AccVideo)
Tianwei Yin (CausVid)
ZuluVision (MoviiGen)
Alibaba PAI (MPS LoRA)
Kijai (ComfyUI Wrapper)

And thanks to the open-source community!