GitHub - harry0703/MoneyPrinterTurbo: 利用 AI 大模型和自动化工作流,根据主题或关键词一键生成高清短视频。Generate HD short videos from a topic or keyword with an automated AI workflow.

Open original

MoneyPrinterTurbo is an open-source project that takes a video topic or keyword and produces a finished short video: script, stock or AI-generated footage, subtitles, voiceover, and background music, all stitched together. Most of the page is sponsor blurbs and install instructions for Windows, Docker, uv, and pip.

  • Generates the full pipeline from one topic: script, voiceover, footage matching, subtitles, music, and editing, with optional custom content at each step.
  • Works with many LLM providers (OpenAI, Claude, Gemini, DeepSeek, Qwen, and others) and many TTS engines, including free Edge TTS that needs no API key.
  • Footage can come from Pexels, Pixabay, Coverr, uploaded files, or AI video models like MiniMax H3 and Seedance; subtitles default to a fast timestamp mode, with local Whisper as an option.
  • Offers four ways to run it: an AI agent via a Skill file, WebUI, REST API, and CLI with batch-file support.

MoneyPrinterTurbo is an open-source tool that takes a video topic or keyword and produces a finished short video: script, search terms for stock footage, subtitles, background music, and the final render at 1080p. You can drive it through a WebUI, a Python CLI, an HTTP API, or an AI agent that reads its Skill doc. The output formats are 9:16 vertical (1080×1920), 16:9 landscape, and 1:1 square. Supported script and voice providers are a long list, but the practical core is a choice of LLM (Kimi/Moonshot, OpenAI, Claude, Gemini, DeepSeek, Qwen, Doubao, and others, or a gateway like OpenRouter or Ollama for local models), a voice engine (Edge TTS is free and needs no key), and a footage source (Pexels and Pixabay are free, Coverr too). It can also generate the clips themselves through AI video APIs such as MiniMax H3, Seedance, or Wan, and can batch-produce several videos with a task history. Most of the page is installation guidance, and the setup choices are where the real decisions sit. The one-click Windows package is the fastest route, but the path must avoid Chinese characters, special characters, and spaces, and you should grab the `.7z` from the Assets section rather than the auto-generated source zip, which lacks `start.bat`. For Python users, the `mpt` package on PyPI is the official one; a different package named `MoneyPrinterTurbo` on PyPI is third-party. Docker is the cleanest isolated option and pulls a prebuilt image from GitHub Container Registry. Hardware expectations are modest: 4 cores and 4 GB RAM will run it if you rely on cloud LLM, TTS, and footage, but a GPU with 4 to 8 GB of VRAM matters if you turn on local `faster-whisper` transcription or heavy batch work. Python 3.11 or newer is required for local installs. Roughly half the article is sponsor copy, mostly API resellers and aggregators pitching discounts and free credits, and it also lists a Chinese MangoDisk disk-cleaning project that has nothing to do with video generation. Strip those out and what remains is a useful, fairly complete open-source pipeline. The sponsor links carry referral bonuses, so treat the pricing claims as marketing until you check them yourself. Before running it, check that your chosen LLM and footage provider keys are set in the WebUI's basic settings, and note that uploaded custom audio is not saved between CLI runs, so you have to pass the file path explicitly each time.

Ask about this article