In this live stream I demo MoneyPrinterTurbo, a free, open source tool that makes a complete video for you. It writes the script, searches for stock footage, generates subtitles, adds a synthetic voiceover, picks background music, and renders the final file. The finished video I opened the stream with was 99.9% made by the tool, with no manual editing.

For the clean, written, step by step version, read my full MoneyPrinterTurbo walkthrough. This post covers what happened in the live demo, including the part that went wrong.

What it does

MoneyPrinterTurbo is a video generation pipeline with six stages: script generation, footage search, subtitle generation, audio synthesis, music selection, and rendering. It works for vertical short form (9:16) and horizontal (16:9) videos. The GitHub repository is harry0703/MoneyPrinterTurbo.

The free setup, in Google Colab

You can run it on your own machine or online. I used the online route, because it needs no local install:

  1. Open the Colab link in the repository's README and run the first cell to install the dependencies.
  2. Create a free ngrok account and copy your auth token. The tool uses ngrok to give you a web address for its interface. I stored my token in an environment file instead of pasting it into the notebook, so it does not show on screen.
  3. Run the deploy cell. It prints a URL, and that URL opens the interface.

Colab sometimes stops web interfaces on the free tier. If your session dies, run the tool locally or with Docker instead. A GPU is not required, but it is recommended for faster batch rendering. It runs on Windows 10 or newer, macOS 11 or newer, and Linux.

The two free API keys you need

  • An LLM provider key. I picked Gemini, because you can get a free key with limited usage. The settings list OpenAI, Qwen, DeepSeek, Gemini, Grok and Ollama. I recommend Gemini 3.5 Flash.
  • A Pexels API key for the stock footage. It is free, and it is what supplies the B-roll. The settings also list other stock sources, such as Pixabay.

Generating a video

Open the basic settings, pick the provider, and paste in your keys. Then type a video subject. For the finished demo I used "how to make money online for Gen Z." Live, I typed "why 90% of AI agent companies will fail."

The tool then writes a script and a list of search keywords with Gemini. After that you choose:

  • Layout: portrait 9:16 or landscape 16:9.
  • Batch size: up to 5 videos in one run.
  • Voice: male or female, plus the accent and dialect. You can also change the speech rate and the volume.
  • Music: a random track, or your own, with its own volume control.
  • Footage: stock clips, or your own MP4, MOV or WebM files.
  • Subtitles: font, color, outline and position.
  • Language: the same video in other languages, for example Chinese, Spanish or Portuguese.

You can also connect it to an AI agent, so videos can be generated by code instead of by clicking.

What went wrong in the live demo

I asked for two horizontal videos, and the render took far longer than I expected. After more than 25 minutes, the tool was still cutting about 318 sub clips, and I had to end the stream before it finished. That is why you only see the pre-made video and not a live result.

Two lessons from that. Render one video at a time while you test, and pick portrait if you only need short form. Batches and horizontal video multiply the processing.

About the "real footage" point

On stream I said that because the tool uses real stock footage and not AI generated video, it has better monetization potential. That is my opinion, not a guarantee. Platforms also have rules about repetitive or mass produced content, so check the current YouTube and TikTok policies before you publish dozens of these.

If you want to build a content workflow around it, start with one video, edit the script yourself, and judge the result before you scale.