<h1 id="scail-2-gguf">SCAIL-2 GGUF</h1><div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/YqYSaz_aKlk?rel=1" start="0"></iframe></div>Rebels GGUF quantizations of SCAIL-2, the end-to-end character-animation / video motion-transfer model (Wan 2.1 14B backbone) from <a target="_blank" rel="ugc" href="https://huggingface.co/zai-org/SCAIL-2">zai-org</a>. These run the SCAIL-2 DiT in ComfyUI at a fraction of the VRAM the full fp16/fp8 weights require.Quantized by <a target="_blank" rel="ugc" href="https://huggingface.co/realrebelai">RealRebelAI</a> · <a target="_blank" rel="ugc" href="https://github.com/RealRebelAI">GitHub</a> · <a target="_blank" rel="ugc" href="https://youtube.com/@RealRebelAI">YouTube</a>⚡ Load with the GGUF Unet Loader (city96's <a target="_blank" rel="ugc" href="https://github.com/city96/ComfyUI-GGUF">ComfyUI-GGUF</a> — <code>Unet Loader (GGUF)</code>). Place the <code>.gguf</code> in <code>ComfyUI/models/unet/</code>.<hr /><h2 id="quant-tiers">Quant tiers</h2><ul><li>Q2_K 6 GB - Smallest, runs on minimal VRAM, expect quality loss</li><li>Q3_K_M 8 GB - Budget tier, better coherence than Q2</li><li>Q4_K_M 10 GB - Recommended daily driver</li><li>Q5_K_M 12 GB - Sweet spot above Q4</li><li>Q6_K 14 GB - Higher fidelity</li><li>Q8_0 17 GB - Closest to fp16</li></ul>The loader memory-maps the model, so a larger file costs disk and streaming time, not resident RAM.<hr /><h2 id="required-files">Required files</h2>Download each of these separately and place them in the listed ComfyUI folder.<h3 id="model">😎 Model</h3><code>ComfyUI/models/unet/</code><a target="_blank" rel="ugc" href="https://huggingface.co/realrebelai/SCAIL-2_GGUF/tree/main">https://huggingface.co/realrebelai/SCAIL-2_GGUF/tree/main</a><h3 id="text-encoder">📝 Text Encoder</h3><code>ComfyUI/models/text_encoders/</code> <a target="_blank" rel="ugc" href="https://huggingface.co/chatpig/encoder/blob/main/umt5_xxl_fp8_e4m3fn_scaled.safetensors">https://huggingface.co/chatpig/encoder/blob/main/umt5_xxl_fp8_e4m3fn_scaled.safetensors</a><h3 id="lora-(lightx2v-stepcfg-distill)">🎛️ LoRA (LightX2V step/cfg distill)</h3><code>ComfyUI/models/loras/</code> <a target="_blank" rel="ugc" href="https://huggingface.co/lightx2v/Wan2.1-I2V-14B-480P-StepDistill-CfgDistill-Lightx2v/blob/main/loras/Wan21_I2V_14B_lightx2v_cfg_step_distill_lora_rank64.safetensors">https://huggingface.co/lightx2v/Wan2.1-I2V-14B-480P-StepDistill-CfgDistill-Lightx2v/blob/main/loras/Wan21_I2V_14B_lightx2v_cfg_step_distill_lora_rank64.safetensors</a><h3 id="sam-3.1-multiplex">🎯 SAM 3.1 Multiplex</h3><code>ComfyUI/models/sam/</code> <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors">https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors</a><h3 id="clip-vision">👁️ CLIP Vision</h3><code>ComfyUI/models/clip_vision/</code> <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors">https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors</a><h3 id="vae">🎨 VAE</h3><code>ComfyUI/models/vae/</code> <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors">https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors</a><h3 id="optional:-scail-2-dpo-lora-(untested)">➕ Optional: SCAIL-2 DPO LoRA (untested)</h3><code>ComfyUI/models/loras/</code> <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/SCAIL-2/blob/main/loras/wan2.1_SCAIL_2_DPO_lora_bf16.safetensors">https://huggingface.co/Comfy-Org/SCAIL-2/blob/main/loras/wan2.1_SCAIL_2_DPO_lora_bf16.safetensors</a><hr /><h2 id="folder-structure">Folder structure</h2><pre><code>ComfyUI/models/
├── unet/
│ └── SCAIL-2-Q4_K_M.gguf ← from this repo
├── text_encoders/
│ └── umt5-xxl-enc-fp8_e4m3fn.safetensors
├── loras/
│ ├── Wan21_I2V_14B_lightx2v_cfg_step_distill_lora_rank64.safetensors
│ └── wan2.1_SCAIL_2_DPO_lora_bf16.safetensors (optional)
├── sam/
│ └── sam3.1_multiplex_fp16.safetensors
├── clip_vision/
│ └── clip_vision_h.safetensors
└── vae/
 └── wan_2.1_vae.safetensors
</code></pre><hr /><h2 id="notes">Notes</h2><ul><li><code>WEIGHT NOT MERGED</code> warning on <code>patch_embedding</code> is harmless. ComfyUI builds a 36-channel patch embedding and concatenates the mask channels at runtime; the model fills them internally. The stored 20-channel weight is expected. Generation proceeds normally.</li><li>The colored mask is a required input even in single-character Animation Mode — don't remove it from the workflow.</li><li>Set width and height explicitly (both divisible by 16; 832×480 is a good 480p start).</li><li>The <code>SCAIL2ColoredMask</code> node may require a recent / nightly ComfyUI build.</li></ul><hr /><h2 id="credits">Credits</h2><ul><li>Model: <a target="_blank" rel="ugc" href="https://huggingface.co/zai-org/SCAIL-2">zai-org / SCAIL-2</a></li><li>GGUF quantization: <a target="_blank" rel="ugc" href="https://huggingface.co/realrebelai">RealRebelAI</a></li><li>GGUF tooling: <a target="_blank" rel="ugc" href="https://github.com/city96/ComfyUI-GGUF">city96 / ComfyUI-GGUF</a></li></ul>

Files with SHA256: 939d5c2feaf50af248040541a7ffcde63a2d23ecfeb3756457feabdcd37dbb58