Update README with inference tool instructions

README.md

@@ -30,13 +30,39 @@ The full bf16 transformer is 32.5 GB — too large for consumer GPUs like the RT
 
 RTX 4090 (Ada Lovelace, SM89) has native FP8 hardware support.
 
-## Usage
+## Inference Tool
 
-This is a drop-in replacement for the original safetensors checkpoint. The model code automatically detects FP8 tensors and upcasts them to bf16 at load time.
+A Gradio UI, CLI, and REST API for running inference are available at **[SanDiegoDude/JoyAI-Image](https://github.com/SanDiegoDude/JoyAI-Image)** — a handy way to run the model until proper ComfyUI integration lands. Features include auto-download from HuggingFace, multiple memory modes, and bitsandbytes quantization for the text encoder.
 
-1. Download this repo into `ckpts_infer/transformer/`
-2. Ensure you also have the VAE and text encoder from the [original repo](https://huggingface.co/jdopensource/JoyAI-Image-Edit) or [safetensors conversion](https://huggingface.co/SanDiegoDude/JoyAI-Image-Edit-Safetensors)
-3. Run inference normally — the loading code handles the FP8→bf16 upcast transparently
+### Quick start
+
+```bash
+git clone https://github.com/SanDiegoDude/JoyAI-Image.git
+cd JoyAI-Image
+python -m venv .venv && source .venv/bin/activate
+pip install -e .
+
+# Models auto-download on first run (FP8 transformer is the default)
+# Default: FP8 transformer + 8-bit text encoder, offload mode
+python app.py
+
+# Minimum VRAM (~13 GB active, fits RTX 4090)
+python app.py --nf4-dit --4bit-vlm
+
+# CLI inference
+python inference.py \
+  --prompt "Turn the plate blue" \
+  --image test_images/test_1.jpg \
+  --output result.png \
+  --steps 18 --guidance-scale 4.0 --seed 42
+
+# Headless REST API (with optional ComfyUI connector node)
+python app.py --headless-api 7500
+```
+
+## Also available
+
+- **[Full bf16 safetensors](https://huggingface.co/SanDiegoDude/JoyAI-Image-Edit-Safetensors)** — original precision, also the source weights for runtime NF4 quantization (~8 GB)
 
 ## Conversion