Upload README.md

README.md

@@ -0,0 +1,98 @@
+---
+license: openrail++
+tags:
+  - image-to-video
+  - wan2.1
+  - finetune
+  - anime
+  - 480p
+model_type: image-to-video
+base_model: Wan2_1-I2V
+---
+
+# Wanime I2V 14B – Anime-Style Image-to-Video (480p)
+
+This is a Wan 2.1 14B image-to-video finetune model focused on anime-style motion and aesthetics at 480p. Based on the original weights released by Civitai creator [Zazc](https://civitai.com/models/1626197?modelVersionId=1840561), this model is locally prepped for use in WanGP with both FP16 and INT8 quantized formats.
+
+---
+
+## 📥 Files
+
+- `Wanime-I2V-14B-480P_fp16_pure.safetensors` – Original FP16 weights for full precision video generation.
+- `Wanime-I2V-14B-480P_quanto_fp16_int8.safetensors` – Quantized INT8 version for faster generation in lower VRAM setups.
+
+---
+
+## 🛠 How to Use (WanGP)
+
+1. Drop both `.safetensors` files into your WanGP `ckpts/` folder:
+   ```
+   D:/pinokio/api/wan.git/app/ckpts/
+   ```
+
+2. Save the following JSON file to:
+   ```
+   D:/pinokio/api/wan.git/app/finetunes/wan_i2v_wanime14B_480p.json
+   ```
+
+3. Sample `finetune JSON`:
+   ```jsonc
+   {
+     "model": {
+       "name": "Wanime I2V 14B 480P",
+       "architecture": "i2v",
+       "description": "Wanime 480p anime-style I2V model, optimized for local WanGP use.",
+       "URLs": [
+         "Wanime-I2V-14B-480P_fp16_pure.safetensors",
+         "Wanime-I2V-14B-480P_quanto_fp16_int8.safetensors"
+       ],
+       "modules": [],
+       "auto_quantize": false
+     },
+     "prompt": "",
+     "negative_prompt": "out of frame, lowres, text, error, cropped, worst quality, jpeg artifacts, ugly, duplicate, poorly drawn hands, malformed limbs, watermark",
+     "resolution": "832x480",
+     "video_length": 81,
+     "num_inference_steps": 8,
+     "guidance_scale": 1.0,
+     "multi_images_gen_type": 0
+   }
+   ```
+
+4. Launch with:
+   ```bash
+   python wgp.py --multiple-images
+   ```
+
+---
+
+## 🧠 About the Model
+
+This model is derived from Zazc's Civitai release “WANime v3 I2V 14B”:
+🔗 https://civitai.com/models/1626197?modelVersionId=1840561
+
+It was originally trained for consistent anime-character motion generation using Wan2.1’s powerful I2V backbone (14B params). This version is cleaned for `.safetensors` compatibility and formatted for WanGP.
+
+---
+
+## ✨ Recommended Prompts
+
+- “An anime girl walking through a field of sunflowers, wide shot, cinematic composition”
+- “Dramatic silhouette of a ninja jumping across rooftops at night, anime lighting”
+- “Side-scrolling shot of a futuristic anime city, neon lights, sunset glow”
+
+---
+
+## 💡 Notes
+
+- INT8 model recommended for RTX 3060 or lower
+- FP16 model performs best on 4090/3090 class GPUs
+- You can convert to BF16 or quantize your own using `--save-quantized`
+
+---
+
+## 🔖 Credits
+
+- Base model: **Wan2.1 I2V 14B**
+- Finetune creator: [Zazc on Civitai](https://civitai.com/models/1626197)
+- Local packaging & WanGP prep: **hdtrnk**