Reading the repo
Opening the box…
Reading the repo
Opening the box…
How to install on Windows
Extract any sound with text prompts. Memory-optimized SAM-Audio with modern UI.
GitHub didn't answer, so this is the saved copy from 1 days ago
Will it work on Windows?
Heads-upBuilding it yourself? It also needs Redis.
The maintainers included install.bat. Run it from the cloned folder after reading it.
.\install.bat
Also: start.bat, install.sh, start.sh
QUICKSTART.md最簡單的方式:Windows 執行
install.bat一次,之後用start.bat啟動、stop.bat關閉。 macOS(Apple Silicon)/ Linux:./install.sh、./start.sh、./stop.sh(Mac 目前用 CPU 運算,較慢)。 以下是手動安裝步驟(Windows / PowerShell)。
# 方法 A:install.bat 下載的離線版(start.bat 會自動啟動)
# 方法 B:Docker
docker-compose up -d
conda create -n audioghost python=3.11 -y
conda activate audioghost
# RTX 20/30/40 系列(CUDA 12.6)
pip install torch==2.9.0+cu126 torchvision==0.24.0+cu126 torchaudio==2.9.0+cu126 --index-url https://download.pytorch.org/whl/cu126 --extra-index-url https://pypi.org/simple
# RTX 50 系列(Blackwell)請改用 cu128
pip install torch==2.9.0+cu128 torchvision==0.24.0+cu128 torchaudio==2.9.0+cu128 --index-url https://download.pytorch.org/whl/cu128 --extra-index-url https://pypi.org/simple
Windows:下載 FFmpeg 8 的 shared build(例如 BtbN 的 ffmpeg-n8.1-latest-win64-lgpl-shared-8.1.zip:https://github.com/BtbN/FFmpeg-Builds/releases/tag/latest ),解壓縮後把 bin 資料夾加入 PATH(install.bat 會自動做)。
目前請避免在 Windows 上用 conda-forge 的 FFmpeg:它的套件裝出兩份不同的 intl-8.dll,會導致載入失敗(#10)。
macOS / Linux:
conda install -c conda-forge "ffmpeg=7" -y
pip install -c backend\constraints.txt "git+https://github.com/facebookresearch/sam-audio.git@bb4c6999d2677c7402360e426afc01ddfad6dce0"
cd backend
pip install -c constraints.txt -r requirements.txt
cd backend
uvicorn main:app --port 8000
conda activate audioghost
cd backend
celery -A workers.celery_app worker --loglevel=info --pool=solo
cd frontend
npm ci
npm run dev
訪問 http://localhost:3000
-tv) checkpoints, full FP32, max chunk lengthpython sam_audio_lite.py song.mp3 "singing voice"
python sam_audio_lite.py interview.mp4 "background music" --model base --quality fast -o out/
python sam_audio_lite.py live.flac "crowd noise" --stereo --start 12.5 --end 16 --format flac
| Variable | Default | Meaning |
|---|---|---|
| AUDIOGHOST_PRECISION | fp16 | Transformer precision when High Quality is off (fp16 or bf16) |
| AUDIOGHOST_PRELOAD_MODEL | - | small / base / large: load a model when the worker starts |
| AUDIOGHOST_DEVICE | auto | Force cuda, cuda:1, cpu, or mps (Apple GPU, experimental; runs in fp32 unless AUDIOGHOST_PRECISION is set) |
| AUDIOGHOST_SAFE_CODEC | 0 | 1 = use SAM-Audio's original (slow, cuDNN-off) codec path |
| AUDIOGHOST_RETENTION_HOURS | 24 | Delete uploads/results older than this (0 = keep) |
| AUDIOGHOST_MAX_UPLOAD_MB | 2048 | Upload size limit |
| AUDIOGHOST_REDIS_URL | redis://localhost:6379/0 | Celery broker / result backend |
| NEXT_PUBLIC_API_BASE | http://localhost:8000 | Where the web UI finds the API |
Then run powershell -ExecutionPolicy Bypass -File .\setup.ps1. It stops at the first error and is safe to run again. Servers and tests aren't started; it lists them at the end. Read it first.
Needs Git: winget install Git.Git
git clone https://github.com/0x0funky/audioghost-ai.git cd audioghost-ai
Python 3
winget install Python.Python.3.12
Needs: Redis
winget install Docker.DockerDesktop docker compose -f docker-compose.yml up -d
In the maintainers' order. Do the README instructions above first — they cover steps outside the terminal.
python sam_audio_lite.py song.mp3 "singing voice" python sam_audio_lite.py interview.mp4 "background music" --model base --quality fast -o out/ python sam_audio_lite.py live.flac "crowd noise" --stereo --start 12.5 --end 16 --format flac
No reports for Windows yet.
GitManual AI Agent reads the 3 key files and writes a step-by-step guide for Windows. Once written, it's saved for everyone.