Update README.md

README.md

@@ -1,3 +1,268 @@
----
-license: apache-2.0
----
+---
+license: apache-2.0
+language:
+- en
+tags:
+- Text-to-Speech
+---
+# Dia
+
+[![Join us on Discord](https://img.shields.io/badge/Discord-Join%20Chat-7289DA?logo=discord)](https://discord.gg/pgdB5YRe)
+[![License](https://img.shields.io/badge/License-Apache_2.0-blue.svg)](LICENSE)
+
+<!-- Add other badges here: PyPI version, Hugging Face model, etc. -->
+
+Dia is a 1.6B parameter speech generation model created by Nari Labs. Dia can generate highly realistic dialogue from a transcript. You can condition the output on audio, enabling emotion and tone control. The model can also produce nonverbal communications like laughter, coughing, clearing throat, etc.
+
+[Demo Page](https://yummy-fir-7a4.notion.site/dia-demo) comparing our model to [ElevenLabs Studio](https://elevenlabs.io/studio) and [Sesame CSM-1B](https://github.com/SesameAILabs/csm).
+
+## Features
+
+- **Python Library:** Core functionalities accessible via `import nari_tts`.
+- **Command-Line Interface (CLI):** Generate audio from text using `scripts/infer.py`.
+- **Gradio Web UI:** Interactive demo interface via `app/app.py`.
+- **Hugging Face Hub Integration:** Load models directly from the Hub (`buttercrab/nari-tts` placeholder).
+- **Docker Support:** Run the CLI or Gradio app in isolated containers.
+  - `Dockerfile.cli`
+  - `Dockerfile.app`
+  - `docker-compose.yml` (for Gradio app)
+- **Audio Prompting:** Guide speech style using an audio prompt (optional).
+
+## Installation
+
+### Prerequisites
+
+- Python >= 3.10
+- PyTorch >= 2.6.0 (check compatibility with your CUDA version if applicable)
+- `uv` (optional, for faster dependency installation: `pip install uv`)
+
+### Installing the Library
+
+**1. From Source (Recommended for Development):**
+
+Clone the repository and install in editable mode:
+
+```bash
+# Clone the repository (replace with your actual repo URL)
+git clone https://github.com/your-username/nari-dialogue.git
+cd nari-dialogue
+
+# Install using pip (uses pyproject.toml)
+pip install -e .
+
+# Or install using uv (faster)
+uv pip install -e .
+```
+
+**2. From PyPI (Once Published):**
+
+```bash
+pip install nari-tts
+```
+
+**3. Installing Dependencies Separately:**
+
+You can install dependencies using `uv` or `pip` with the provided files:
+
+```bash
+# Using uv (recommended)
+uv pip install -r requirements.txt
+
+# Using pip
+pip install -r requirements.txt
+```
+
+## Usage
+
+### As a Python Library
+
+```python
+import torch
+import dac
+import soundfile as sf
+from pathlib import Path
+from nari_tts import load_model_from_hub, generate, codebook_to_audio
+
+# --- Configuration ---
+REPO_ID = \"buttercrab/nari-tts\" # Replace with actual Hub ID
+DAC_MODEL_TYPE = \"44khz\"
+OUTPUT_FILENAME = \"library_output.wav\"
+INPUT_TEXT = \"This audio was generated using the Nari TTS library.\"
+
+# --- Device ---
+device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")
+print(f\"Using device: {device}\")
+
+# --- Load Models ---
+print(f\"Loading Nari model from {REPO_ID}...\")
+model, config = load_model_from_hub(REPO_ID, device=device)
+
+print(f\"Loading DAC model ({DAC_MODEL_TYPE})...\")
+dac_model_path = dac.utils.download(model_type=DAC_MODEL_TYPE)
+dac_model = dac.DAC.load(dac_model_path).to(device)
+dac_model.eval()
+dac_sample_rate = dac_model.sample_rate
+
+# --- Generate ---
+print(f\"Generating audio for: '{INPUT_TEXT}'\")
+generated_codes = generate(
+    model=model,
+    config=config,
+    text=INPUT_TEXT,
+    max_tokens=config.data.audio_length, # Or specify a number
+    cfg_scale=3.0,
+    temperature=1.0,
+    top_p=0.95,
+    use_cfg_filter=True,
+    device=device,
+    cfg_filter_top_k=50,
+    # audio_prompt_path=\"path/to/prompt.wav\", # Optional
+    dac_model=dac_model
+)
+
+# --- Decode and Save ---
+if generated_codes.numel() > 0:
+    print(\"Converting codes to audio...\")
+    audio_tensor = codebook_to_audio(
+        generated_codes=generated_codes.transpose(0, 1), # Needs [C, T]
+        dac_model=dac_model,
+        delay_pattern=config.data.delay_pattern,
+        C=config.data.channels,
+    )
+    audio_np = audio_tensor.cpu().float().numpy().squeeze()
+    sf.write(OUTPUT_FILENAME, audio_np, dac_sample_rate)
+    print(f\"Audio saved to {OUTPUT_FILENAME}\")
+else:
+    print(\"Generation failed.\")
+
+```
+
+### Command-Line Interface (CLI)
+
+The CLI script `scripts/infer.py` allows generation from the terminal.
+
+**Basic Usage (Loading from Hub):**
+
+```bash
+python scripts/infer.py \"Your input text goes here.\" \
+    --repo-id buttercrab/nari-tts \
+    --output generated_speech.wav
+```
+
+**Loading from Local Files:**
+
+```bash
+python scripts/infer.py \"Text for local model.\" \
+    --local-paths \
+    --config path/to/your/config.json \
+    --checkpoint path/to/your/nari_v0.pth \
+    --output local_output.wav
+```
+
+**With Audio Prompt:**
+
+```bash
+python scripts/infer.py \"Generate speech like this prompt.\" \
+    --repo-id buttercrab/nari-tts \
+    --audio-prompt path/to/your/prompt.wav \
+    --output prompted_output.wav
+```
+
+**See all options:**
+
+```bash
+python scripts/infer.py --help
+```
+
+### Gradio Web UI
+
+The Gradio app provides an interactive interface.
+
+**Running Locally:**
+
+```bash
+# Load from Hub (replace repo-id)
+python app/app.py --repo-id buttercrab/nari-tts
+
+# Load from local files
+python app/app.py --local-paths --config path/to/config.json --checkpoint path/to/checkpoint.pth
+```
+
+Access the UI in your browser (usually at `http://127.0.0.1:7860`).
+
+### Docker
+
+**1. Build the Images:**
+
+```bash
+# Build CLI image
+docker build -t nari-tts-cli:latest -f Dockerfile.cli .
+
+# Build App image
+docker build -t nari-tts-app:latest -f Dockerfile.app .
+```
+
+**2. Run CLI Container:**
+
+```bash
+docker run --rm -v \"$(pwd)/output:/app/output\" nari-tts-cli:latest \
+    \"Generating audio inside a Docker container.\" \
+    --repo-id buttercrab/nari-tts \
+    --output /app/output/docker_output.wav
+
+# Mount ~/.cache/huggingface to reuse downloads:
+# docker run --rm -v \"$(pwd)/output:/app/output\" -v \"~/.cache/huggingface:/root/.cache/huggingface\" nari-tts-cli:latest ...
+```
+
+_(Note: Adjust volume mounts as needed for your OS and paths.)_
+
+**3. Run Gradio App Container (using Docker Compose):**
+
+This is the easiest way to run the Gradio app with proper port mapping and volume mounts.
+
+```bash
+docker compose up
+```
+
+This will build the `nari-tts-app` image if it doesn't exist and start the container.
+Access the UI at `http://localhost:7860`.
+
+To stop the service:
+
+```bash
+docker compose down
+```
+
+## Model
+
+_(Add details about the model architecture, training data, expected quality, and link to the Hugging Face model card here)_.
+
+## Disclaimer
+
+This project offers a high-fidelity speech generation model intended solely for research and educational use. The following uses are **strictly forbidden**:
+
+- **Identity Misuse**: Do not produce audio resembling real individuals without permission.
+- **Deceptive Content**: Do not use this model to generate misleading content (e.g. fake news)
+- **Illegal or Malicious Use**: Do not use this model for activities that are illegal or intended to cause harm.
+
+By using this model, you agree to uphold relevant legal standards and ethical responsibilities. We **are not responsible** for any misuse and firmly oppose any unethical usage of this technology.
+
+## License
+
+This project is licensed under the Apache License 2.0 - see the [LICENSE](LICENSE) file for details.
+
+## TODO / Future Work
+
+- Optimize inference speed.
+- Add quantization for memory efficiency.
+
+## Contributing
+
+We are a tiny team of 1 full-time and 1 part-time research-engineers. We are extra-welcome to any contributions!
+Join our [Discord Server](https://discord.gg/pgdB5YRe) for discussions.
+
+## Acknowledgements
+
+- We thank the [Google TPU Research Cloud program](https://sites.research.google/trc/about/) for providing computation resources.
+- Our work was heavily inspired by [SoundStorm](https://arxiv.org/abs/2305.09636) and [Parakeet](https://jordandarefsky.com/blog/2024/parakeet/).
+- "Nari" is a pure Korean word for lily.