Upload README.md with huggingface_hub

README.md

@@ -9,209 +9,86 @@ tags:
 - lora
 - multi-angle
 - camera-angles
-- 3d-poses
 - text-to-image
-- image-generation
 - fal-ai
 base_model: black-forest-labs/FLUX.1-dev
 pipeline_tag: text-to-image
-widget:
-- text: "<sks> front view eye-level shot close-up, a robot"
-  output:
-    url: poses_72_animated.gif
 ---
 
-# 🎬 Flux.2 Multi-Angles v2 - 72 Camera Poses LoRA
+# 🎬 Flux 2 Multi-Angles LoRA v2
 
-> **Generate images from any angle with precise camera control!**
+> **Control camera angles with 72 unique positions - Trained with [fal.ai](https://fal.ai)**
 
-<p align="center">
-  <a href="https://fal.ai"><img src="https://fal.ai/favicon.ico" width="20" height="20" /> <strong>Trained with fal.ai</strong></a>
-</p>
+## 🖼️ Results
 
-![72 Poses Animated](poses_72_animated.gif)
+| Input → Output | Input → Output | Input → Output |
+|----------------|----------------|----------------|
+| ![Result 1](result01.png) | ![Result 2](result02.png) | ![Result 3](result03.png) |
 
-This LoRA enables **72 unique camera positions** for FLUX.2, giving you complete control over viewing angles, elevations, and distances.
+![72 Poses Demo](poses_72_animated.gif)
 
 ---
 
-## 🎯 Quick Start - Prompt Format
+## 🎯 Prompt Format
 
 ```
 <sks> [view] [elevation] shot [distance]
 ```
 
-### Examples
+### Quick Examples
 
 ```
 <sks> front view eye-level shot close-up
 <sks> right side view high-angle shot medium shot
 <sks> back view overhead shot wide shot
-<sks> front-left quarter view mid-angle shot close-up
 ```
 
-### Complete Prompt Reference
+### All 72 Prompts Reference
 
 ![72 Poses Prompts](poses_72_prompts.png)
 
 ---
 
-## 📐 Camera Configuration
-
-### 8 Azimuth Angles (Horizontal)
-| Angle | View Name |
-|-------|-----------|
-| 0° | front view |
-| 45° | front-right quarter view |
-| 90° | right side view |
-| 135° | back-right quarter view |
-| 180° | back view |
-| 225° | back-left quarter view |
-| 270° | left side view |
-| 315° | front-left quarter view |
-
-### 9 Elevation Angles (Vertical)
-| Angle | Shot Type |
-|-------|-----------|
-| 0° | eye-level shot |
-| 10° | low-angle shot |
-| 20° | mid-low shot |
-| 30° | mid-angle shot |
-| 40° | high-mid shot |
-| 45° | high-angle shot |
-| 50° | steep-mid shot |
-| 60° | steep-angle shot |
-| 75° | overhead shot |
-
-### 3 Distances (Depth)
-| Distance | Radius |
-|----------|--------|
-| close-up | 3.0 |
-| medium shot | 4.0 |
-| wide shot | 5.0 |
-
-![72 Poses Hemisphere](poses_72_hemisphere.png)
+## 📦 Files
 
----
-
-## 📦 Files Included
-
-| File | Description | Use Case |
-|------|-------------|----------|
-| `flux-multi-angles-v2-72poses-fal.safetensors` | FAL-compatible LoRA | Use with [fal.ai](https://fal.ai) API |
-| `flux-multi-angles-v2-72poses-comfy.safetensors` | ComfyUI-compatible LoRA | Use with ComfyUI |
+| File | Use Case |
+|------|----------|
+| `flux-multi-angles-v2-72poses-fal.safetensors` | [fal.ai](https://fal.ai) |
+| `flux-multi-angles-v2-72poses-comfy.safetensors` | ComfyUI |
 
 ---
 
-## 🚀 Usage
-
-### With fal.ai
-
-This model was trained using [**fal.ai**](https://fal.ai) infrastructure.
-
-1. Use `flux-multi-angles-v2-72poses-fal.safetensors`
-2. Upload to FAL or reference from Hugging Face
-3. Use the prompt format below
-
-### With ComfyUI
-
-1. Download `flux-multi-angles-v2-72poses-comfy.safetensors`
-2. Place in your `ComfyUI/models/loras/` folder
-3. Load FLUX.2-dev as base model
-4. Apply LoRA with strength 0.8-1.0
-5. Use the prompt format: `<sks> [view] [elevation] shot [distance], [your subject]`
-
-### Example Prompts
-
-```python
-# Front view, eye level, close-up
-"<sks> front view eye-level shot close-up, a futuristic robot"
-
-# Side view, high angle, medium distance
-"<sks> right side view high-angle shot medium shot, a sports car"
-
-# Back view, overhead, wide shot
-"<sks> back view overhead shot wide shot, a medieval castle"
-```
-
----
-
-## ⚙️ Recommended Settings
+## ⚙️ Settings
 
 - **LoRA Strength**: 0.8 - 1.0
-- **CFG Scale**: 3.5 - 7.0
-- **Steps**: 20-30
 - **Base Model**: FLUX.2-dev
 
 ---
 
-## 🆕 What's New in v2?
-
-| Feature | v1 (24 Poses) | v2 (72 Poses) |
-|---------|---------------|---------------|
-| Azimuths | 8 | 8 |
-| Elevations | 3 | 9 |
-| Distances | 1 (close-up) | 3 (close-up, medium, wide) |
-| **Total Poses** | **24** | **72** |
-
-![24 vs 72 Comparison](correct_24_vs_72_comparison.png)
-
----
-
-## 📊 Dataset Creation Methodology
+## 📐 72 Camera Positions
 
-This LoRA was trained with [**fal.ai**](https://fal.ai) using a carefully constructed dataset featuring:
+**8 Azimuths** × **9 Elevations** × **3 Distances** = **72 Poses**
 
-- **72 precise camera positions** mathematically calculated on a hemisphere
-- **Consistent subjects** rendered from all angles
-- **Clear labeling system** using view + elevation + distance format
-- **Multiple distance variations** to teach depth perception
+| Azimuths | Elevations | Distances |
+|----------|------------|-----------|
+| front, front-right quarter, right side, back-right quarter, back, back-left quarter, left side, front-left quarter | eye-level, low-angle, mid-low, mid-angle, high-mid, high-angle, steep-mid, steep-angle, overhead | close-up, medium, wide |
 
-The goal was to create a **reproducible and documented approach** that others can learn from and build upon.
+![Camera Positions](poses_72_hemisphere.png)
 
 ---
 
-## 💡 Inspiration & Background
-
-This project was inspired by two existing works:
+## 💡 Inspiration
 
-### 1. My Previous Work: Next-Scene LoRA
-I previously created [**next-scene-qwen-image-lora-2509**](https://huggingface.co/lovis93/next-scene-qwen-image-lora-2509) - a LoRA for Qwen-Image-Edit that enables cinematic scene progression and camera movements. This gave me experience in training models for camera control.
+This project was inspired by:
+- My previous work [next-scene-qwen-image-lora](https://huggingface.co/lovis93/next-scene-qwen-image-lora-2509)
+- [dx8152/Qwen-Edit-2509-Multiple-angles](https://huggingface.co/dx8152/Qwen-Edit-2509-Multiple-angles)
 
-### 2. dx8152's Multiple-Angles LoRA
-I was also inspired by [**dx8152/Qwen-Edit-2509-Multiple-angles**](https://huggingface.co/dx8152/Qwen-Edit-2509-Multiple-angles) - a similar concept for camera angle control. However, since the dataset creation process wasn't documented, **I wanted to demonstrate how to properly create a dataset for this type of model** and explain the methodology behind it.
-
-### 🎯 My Goal
-With this project, I aimed to:
-- **Show the complete dataset creation process** - transparent methodology
-- **Document the 72-pose system** - precise mathematical angles
-- **Provide comprehensive prompt references** - easy-to-follow examples
-- **Apply these concepts to Flux.2** - bringing multi-angle control to a different architecture
-
----
-
-## 📄 License
-
-This model is released under the [CreativeML Open RAIL-M License](https://huggingface.co/spaces/CompVis/stable-diffusion-license).
+I wanted to document how to properly create a dataset for camera angle control and apply it to Flux.2.
 
 ---
 
 ## 👤 Author
 
-Created by **lovis93** ([@lovis93](https://huggingface.co/lovis93))
-
-### Other Projects:
-- 🎥 [next-scene-qwen-image-lora-2509](https://huggingface.co/lovis93/next-scene-qwen-image-lora-2509) - Cinematic scene progression for Qwen-Image-Edit (521+ likes!)
-
----
-
-## 🙏 Acknowledgments
-
-- [**fal.ai**](https://fal.ai) for providing the training infrastructure
-- Black Forest Labs for FLUX.2
-- [dx8152](https://huggingface.co/dx8152) for the inspiration with [Qwen-Edit-2509-Multiple-angles](https://huggingface.co/dx8152/Qwen-Edit-2509-Multiple-angles)
-- The Hugging Face community
-
----
+Created by **[@lovis93](https://huggingface.co/lovis93)** • Trained with [**fal.ai**](https://fal.ai)
 
-**If you find this model useful, please consider giving it a ⭐!**
+**⭐ If you find this useful, please star!**