HoloCine - Wan 2.2 - Alpha

Details

Download Files

Model description

# WAN 2.2 / HoloCine – Quick Pull Links & Notes (use WAN 2.1 Vae, just tested 2.2 and it fails.)

Pipeline note: FP8 / Scaled T2V models require the UMT5 FP16 text encoder below.

(FP8 text encoders will not work very well.)

---

## Downloads

### HoloCine (T2V, FP8 Scaled)

- https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/T2V/HoloCine/Wan2_2-T2V-A14B-LOW-HoloCine-full_fp8_e4m3fn_scaled_KJ.safetensors

- https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/T2V/HoloCine/Wan2_2-T2V-A14B-HIGH-HoloCine-full_fp8_e4m3fn_scaled_KJ.safetensors

### Required Text Encoder (FP16)

MUST HAVE

https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors

---

### HoloCine — GGUF (Optional)

- https://huggingface.co/QuantStack/HoloCine-GGUF/resolve/main/Full/HighNoise/HoloCine-Full-HighNoise-Q8_0.gguf

- https://huggingface.co/QuantStack/HoloCine-GGUF/resolve/main/Full/LowNoise/HoloCine-Full-LowNoise-Q8_0.gguf

---

### LoRAs (LightSpeed & Fun-Reward) — Grab All for Combo Testing

- https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_t2v_lightx2v_4steps_lora_v1.1_low_noise.safetensors

- https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_t2v_lightx2v_4steps_lora_v1.1_high_noise.safetensors

- https://huggingface.co/alibaba-pai/Wan2.2-Fun-Reward-LoRAs/resolve/main/Wan2.2-Fun-A14B-InP-high-noise-MPS.safetensors

- https://huggingface.co/alibaba-pai/Wan2.2-Fun-Reward-LoRAs/resolve/main/Wan2.2-Fun-A14B-InP-low-noise-MPS.safetensors

- https://huggingface.co/alibaba-pai/Wan2.2-Fun-Reward-LoRAs/resolve/main/Wan2.2-Fun-A14B-InP-low-noise-HPS2.1.safetensors

- https://huggingface.co/alibaba-pai/Wan2.2-Fun-Reward-LoRAs/resolve/main/Wan2.2-Fun-A14B-InP-high-noise-HPS2.1.safetensors

---

## Speed / Quality Notes

Sweet spot: 512×512 frames, length 220–250.

Roughly 5–8 minutes per 15 s clip on an H200 (GPU I/O dependent).

Example progress logs:

0%| | 0/10 [00:00<?, ?it/s] got prompt

50%|█████ | 5/10 [01:14<01:16, 15.25s/it]

^^^ 480x480 | 241 length ^^^

got prompt

20%|██ | 2/10 [01:43<06:47, 50.92s/it] got prompt

30%|███ | 3/10 [02:37<06:08, 52.65s/it]

^^^ 480x832 | 205 length ^^^

---

## Suggested Combos

- High model: HoloCine Scaled High

- Low model: T2V Low – A14

- Swap LoRA High/Low Reward pairs as needed — some scenes respond better to inverted noise mixes.

---

## Prompt Samples (18+ Safe)

### Late-Night Fridge Scene

[global caption]

The scene features [character1], an adult woman with medium-length blonde hair and soft makeup.

It’s 2 AM, and she’s raiding the fridge. [mood: relaxed, playful] [lighting: cool fridge light, soft bounce] [camera: cinematic handheld]

This scene contains 3 shots.

[shot cut] Close-up of [character1] opening the fridge, bathed in blue light.

[shot cut] Medium shot – she scoops ice cream with a smile.

[shot cut] Wide shot – she sits on the kitchen floor with snacks around her, content.

---

### Vanity Mirror Scene

[global caption]

The scene features [character1], an adult woman with medium-length blonde hair and subtle makeup.

[setting: bedroom vanity, soft afternoon light] [mood: confident, playful] [camera: cinematic handheld]

This scene contains 3 shots.

[shot cut] Medium shot – [character1] adjusts her hair in the mirror, smiling. [lighting: warm daylight] [motion: slight camera sway]

[shot cut] Close-up – [character1] fastens a silver necklace, eyes focused, reflection visible. [depth of field: shallow]

[shot cut] Wide shot – [character1] twirls once in front of the mirror and walks off frame. [camera: static tripod]

---

### Template for New Scenes

[global caption] The scene features [character1], [short descriptor]. [setting: ___] [mood: ___] [lighting: ___]

(optional additional adult character info)

This scene contains N shots.

[shot cut] [camera angle], [action], [extra tags...]

[shot cut] ...

---

Your girl, ~~~Nexa~~~

Switch LoRA High/Low Fun-Reward as needed • Tune CFG / flow / denoise per pipeline

Images made by this model

No Images Found.