{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# AI-RVC 一键 AI 翻唱 (Colab)\n",
    "\n",
    "[Open in Colab](https://colab.research.google.com/github/mason369/AI-RVC/blob/master/AI_RVC_Colab.ipynb)\n",
    "\n",
    "这是 AI-RVC 的 Google Colab 运行入口。Notebook 会创建独立 Python 3.10 环境，再运行项目安装脚本，避免 Colab 默认 Python 版本变化影响 `fairseq==0.12.2`。\n",
    "\n",
    "当前处理链路模型：\n",
    "\n",
    "| 环节 | 默认模型 |\n",
    "|------|----------|\n",
    "| 人声分离 | `ensemble:vocal_rvc` |\n",
    "| 卡拉OK分离 | `ensemble:karaoke` |\n",
    "| 去混响 / DeEcho | `dereverb_mel_band_roformer_anvuew_sdr_19.1729.ckpt` |\n",
    "| 内容特征 | `hubert_base.pt` |\n",
    "| F0 音高 | `rmvpe.pt` |\n",
    "| VC 推理 | RVC v2 `.pth` + 可选 FAISS `.index` |\n",
    "\n",
    "使用前请在 Colab 菜单选择：`代码执行程序 -> 更改运行时类型 -> T4 GPU`，然后按顺序运行每个单元格。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. 检查 Colab 运行时\n",
    "\n",
    "确认当前机器有 GPU，并显示 Colab 默认 Python。后续实际安装会使用独立 Python 3.10 环境。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "import subprocess\n",
    "\n",
    "print('Colab 默认 Python:', sys.version)\n",
    "print('\\nGPU 信息:')\n",
    "subprocess.run(['nvidia-smi'], check=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. 克隆或刷新仓库\n",
    "\n",
    "如果 `/content/AI-RVC` 已存在，会重置到 GitHub `master` 分支，确保 Colab 环境和仓库最新代码一致。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%bash\n",
    "set -euo pipefail\n",
    "cd /content\n",
    "if [ -d AI-RVC/.git ]; then\n",
    "  git -C AI-RVC fetch origin master\n",
    "  git -C AI-RVC reset --hard origin/master\n",
    "else\n",
    "  git clone https://github.com/mason369/AI-RVC.git AI-RVC\n",
    "fi\n",
    "cd /content/AI-RVC\n",
    "git rev-parse --short HEAD\n",
    "ls -la"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. 创建 Python 3.10 环境并安装依赖\n",
    "\n",
    "这个步骤会安装 `uv`，用它准备 Python 3.10，然后调用项目自带 `install.py --no-run`。如果任何依赖安装失败，单元格会直接停止并显示错误。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%bash\n",
    "set -euo pipefail\n",
    "cd /content/AI-RVC\n",
    "\n",
    "sudo apt-get update -y\n",
    "sudo apt-get install -y ffmpeg build-essential python3-dev curl git\n",
    "\n",
    "if ! command -v uv >/dev/null 2>&1; then\n",
    "  curl -LsSf https://astral.sh/uv/install.sh | sh\n",
    "fi\n",
    "export PATH=\"$HOME/.local/bin:$PATH\"\n",
    "uv --version\n",
    "uv python install 3.10\n",
    "uv venv --seed --python 3.10 venv310\n",
    "PY=/content/AI-RVC/venv310/bin/python\n",
    "$PY --version\n",
    "$PY -m pip --version\n",
    "$PY -m pip install --upgrade pip\n",
    "$PY -m pip install torch torchaudio --index-url https://download.pytorch.org/whl/cu126\n",
    "$PY - <<'PY'\n",
    "import torch\n",
    "print('Torch:', torch.__version__)\n",
    "print('CUDA available:', torch.cuda.is_available())\n",
    "if not torch.cuda.is_available():\n",
    "    raise RuntimeError('GPU PyTorch install completed but CUDA is not available.')\n",
    "print('CUDA:', torch.version.cuda)\n",
    "print('GPU:', torch.cuda.get_device_name(0))\n",
    "PY\n",
    "uv run --python 3.10 python install.py --no-run"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4. 严格验证依赖与默认模型配置\n",
    "\n",
    "这里会确认关键包、Python 版本、CUDA、Colab 默认分离模型、卡拉OK模型、DeEcho模型都符合当前项目配置。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%bash\n",
    "set -euo pipefail\n",
    "cd /content/AI-RVC\n",
    "PY=/content/AI-RVC/venv310/bin/python\n",
    "$PY - <<'PY'\n",
    "import importlib.metadata as md\n",
    "import sys\n",
    "import torch\n",
    "from infer.separator import (\n",
    "    ROFORMER_DEFAULT_MODEL,\n",
    "    KARAOKE_DEFAULT_MODEL,\n",
    "    ROFORMER_DEREVERB_DEFAULT_MODEL,\n",
    ")\n",
    "\n",
    "assert sys.version_info[:2] == (3, 10), sys.version\n",
    "print('Python:', sys.version)\n",
    "print('Torch:', torch.__version__)\n",
    "print('CUDA available:', torch.cuda.is_available())\n",
    "if not torch.cuda.is_available():\n",
    "    raise RuntimeError('Colab runtime did not provide CUDA. Please switch runtime type to T4 GPU and rerun.')\n",
    "\n",
    "expected = {\n",
    "    'audio-separator': '0.44.1',\n",
    "    'fairseq': '0.12.2',\n",
    "    'gradio': '3.50.2',\n",
    "}\n",
    "for dist, minimum in expected.items():\n",
    "    version = md.version(dist)\n",
    "    print(f'{dist}: {version}')\n",
    "\n",
    "assert ROFORMER_DEFAULT_MODEL == 'ensemble:vocal_rvc', ROFORMER_DEFAULT_MODEL\n",
    "assert KARAOKE_DEFAULT_MODEL == 'ensemble:karaoke', KARAOKE_DEFAULT_MODEL\n",
    "assert ROFORMER_DEREVERB_DEFAULT_MODEL == 'dereverb_mel_band_roformer_anvuew_sdr_19.1729.ckpt', ROFORMER_DEREVERB_DEFAULT_MODEL\n",
    "\n",
    "import faiss  # noqa: F401\n",
    "from audio_separator.separator import Separator  # noqa: F401\n",
    "import fairseq  # noqa: F401\n",
    "print('Dependency and model default checks passed.')\n",
    "PY"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 5. 下载基础模型\n",
    "\n",
    "下载 HuBERT、RMVPE、UVR5 HP2 等必需模型。模型会保存到仓库的 `assets/` 目录。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%bash\n",
    "set -euo pipefail\n",
    "cd /content/AI-RVC\n",
    "PY=/content/AI-RVC/venv310/bin/python\n",
    "$PY tools/download_models.py\n",
    "$PY - <<'PY'\n",
    "from tools.download_models import REQUIRED_MODELS, check_model\n",
    "missing = [name for name in REQUIRED_MODELS if not check_model(name)]\n",
    "if missing:\n",
    "    raise RuntimeError('Missing required models: ' + ', '.join(missing))\n",
    "print('Required models are present:', ', '.join(REQUIRED_MODELS))\n",
    "PY"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 6. 可选：查看或下载声线模型\n",
    "\n",
    "处理链路模型已经在前面准备好。这里的 `.pth` 声线模型是 RVC 目标音色，不属于分离 / F0 / HuBERT 等处理模型。可以先下载一个示例模型，也可以跳过后在 Web UI 中下载。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%bash\n",
    "set -euo pipefail\n",
    "cd /content/AI-RVC\n",
    "PY=/content/AI-RVC/venv310/bin/python\n",
    "$PY - <<'PY'\n",
    "from tools.character_models import list_available_characters, list_available_series\n",
    "print('Available series:')\n",
    "for series in list_available_series():\n",
    "    print(' -', series)\n",
    "chars = list_available_characters()\n",
    "print('\\nFirst 20 voice models:')\n",
    "for char in chars[:20]:\n",
    "    print(f\" - {char['name']}: {char.get('display', char.get('description', ''))}\")\n",
    "print('\\nTotal:', len(chars))\n",
    "PY"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Optional: download one example voice model.\n",
    "# Change character_name as needed, then run this cell.\n",
    "\n",
    "# %%bash\n",
    "# set -euo pipefail\n",
    "# cd /content/AI-RVC\n",
    "# PY=/content/AI-RVC/venv310/bin/python\n",
    "# $PY - <<'PY'\n",
    "# from tools.character_models import download_character_model\n",
    "# character_name = 'rin'\n",
    "# if not download_character_model(character_name):\n",
    "#     raise RuntimeError(f'Failed to download voice model: {character_name}')\n",
    "# print(f'Downloaded voice model: {character_name}')\n",
    "# PY"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 7. 全模式处理矩阵验证\n",
    "\n",
    "上传一段含有人声的短音乐文件到 `/content/input_audio.*`，或把自己的文件路径填到 `INPUT_AUDIO`。此步骤会下载示例声线模型 `rin`，然后依次验证 RoFormer、Karaoke、Demucs、UVR5、官方 1:1、官方唱歌修复等处理模式。任何模式失败都会停止并在 JSON 里保留 traceback。\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Upload a short vocal music file, then set INPUT_AUDIO to its path.\n",
    "# Example after upload: INPUT_AUDIO = '/content/input_audio.wav'\n",
    "INPUT_AUDIO = '/content/input_audio.wav'\n",
    "\n",
    "from pathlib import Path\n",
    "if not Path(INPUT_AUDIO).exists():\n",
    "    raise FileNotFoundError(\n",
    "        f'Missing test audio: {INPUT_AUDIO}. Upload a short vocal music file to this path before running the matrix.'\n",
    "    )\n",
    "print('Matrix input:', INPUT_AUDIO, Path(INPUT_AUDIO).stat().st_size, 'bytes')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%bash\n",
    "set -euo pipefail\n",
    "cd /content/AI-RVC\n",
    "PY=/content/AI-RVC/venv310/bin/python\n",
    "$PY - <<'PY'\n",
    "from tools.character_models import download_character_model\n",
    "if not download_character_model('rin'):\n",
    "    raise RuntimeError('Failed to download voice model: rin')\n",
    "print('Voice model ready: rin')\n",
    "PY\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%bash\n",
    "set -euo pipefail\n",
    "cd /content/AI-RVC\n",
    "PY=/content/AI-RVC/venv310/bin/python\n",
    "$PY tools/run_mode_matrix.py \\\n",
    "  --input /content/input_audio.wav \\\n",
    "  --output-dir /content/AI-RVC/outputs/mode_matrix_colab \\\n",
    "  --require-cuda\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 8. 启动 Web UI\n",
    "\n",
    "运行后 Colab 会输出 Gradio 公共链接。打开链接即可使用 AI-RVC Web UI。此单元格会再次检查环境和必需模型。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%bash\n",
    "set -euo pipefail\n",
    "cd /content/AI-RVC\n",
    "PY=/content/AI-RVC/venv310/bin/python\n",
    "$PY run.py --host 0.0.0.0 --port 7860 --share"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 常见问题\n",
    "\n",
    "| 问题 | 处理方式 |\n",
    "|------|----------|\n",
    "| `nvidia-smi` 失败 | 在 Colab 菜单选择 GPU 运行时后重新运行 |\n",
    "| `fairseq` 安装失败 | 确认本 notebook 的第 3 步使用了 Python 3.10 环境 |\n",
    "| `CUDA out of memory` | 换短音频，或在 UI 中降低处理负载 / 切换分离器 |\n",
    "| 模型下载失败 | 重新运行第 5 步；如果 Hugging Face 限流，可稍后再试或配置 token |\n",
    "\n",
    "本项目仅供学习研究和个人娱乐用途，请勿用于商业用途、欺诈、冒充他人或侵犯权益的场景。"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "gpuType": "T4",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "name": "python3"
  },
  "language_info": {
   "name": "python"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}