A video assembly line, not a video model
MoneyPrinterTurbo takes a topic or a keyword and returns a finished vertical or horizontal short video: it writes the script with a language model, pulls matching stock clips, generates subtitles, synthesizes a voiceover, and lays in background music, then composites the lot into 1080x1920 or 1920x1080. The name signals the intent plainly. This is built for cranking out TikTok and Shorts content at volume.
The important mental model, and the thing the feature list does not say outright, is that this is an orchestrator rather than a generative video model. It does not dream up footage. It stitches together a language model’s words, a stock library’s clips, and a text-to-speech voice. Understanding that tells you exactly where it shines and where it hits a wall.
What it actually wires together
- A script step that calls a model provider. The README lists a long roster including OpenAI, Azure, Gemini, DeepSeek, Moonshot, Qwen, ERNIE, MiniMax, and a local Ollama option, so you are not locked to one vendor.
- A footage step that sources high-definition, copyright-free clips, with Pexels as the configured provider, or your own local material.
- Subtitle generation with control over font, position, color, size, and stroke.
- Multiple synthesized voices with live preview, plus background music with volume control.
- Batch generation, so you produce several candidates and keep the best, and both a Web UI and an API behind a clean MVC layout.
Install
Clone and run from source, the path the maintainer treats as primary:
git clone https://github.com/harry0703/MoneyPrinterTurbo.gitCopy config.example.toml to config.toml and fill in pexels_api_keys and your llm_provider with its key. On macOS and Linux the recommended setup is uv sync --frozen; for an isolated environment, Docker is the documented route. Windows users get a one-click portable bundle, with one caveat worth heeding below.
The gotchas the issue history exposes
The README is a tour of features. The closed issues, the ones that drew real discussion, are where the friction lives:
- The voiceover step is the fragile link. Multiple high-comment threads are TTS failures: errors during audio generation and 403 responses from the voice service. If a run dies, this is the first place to look.
- Compositing has bitten people. A well-discussed thread reports
combine_videosnot working at all in some setups, which is the final and most painful place for a pipeline to fail. - The Windows bundle lags the code. The README itself flags that the packaged Windows starter is an older v1.2.6 build and tells you to run
update.batbeforestart.bat. Skipping that is a common early stumble. - Networking assumptions. The docs note that a VPN may need global traffic mode, and that non-ASCII paths cause unpredictable failures. Both are easy to trip over outside the maintainer’s environment.
None of these are dealbreakers, and the open-issue count is low (30 as of 2026-06), which suggests issues get triaged. But this is a chain of external services, and a chain breaks at its weakest call.
Where it fits, and the ceiling
Reach for MoneyPrinterTurbo when you want faceless, narrated, stock-footage shorts at scale: explainer clips, listicles, motivational reels. The batch mode and provider flexibility make it a genuine production tool for that lane.
The ceiling is the footage. Because the visuals are keyword-matched stock clips rather than generated scenes, the video is only as on-topic as the library allows, and abstract or niche scripts often pull loosely related B-roll. If you need footage that actually depicts your specific script, this architecture cannot get you there. That is a property of the assembly-line design, not a bug to wait out.
MoneyPrinterTurbo versus its lineage and neighbors
| MoneyPrinterTurbo | MoneyPrinter | ShortGPT | |
|---|---|---|---|
| Stars | 84,230 | 13,426 | 7,397 |
| Interface | Web UI and API | script | framework |
| License | MIT | MIT | MIT |
| Best at | batch shorts with a UI | the minimal original | programmable pipelines |
Counts are from GitHub as of June 2026. MoneyPrinter is the leaner original this project’s name nods to. ShortGPT leans toward a programmable framework for automated content. MoneyPrinterTurbo’s edge is the polished Web UI, batch generation, and the breadth of model providers it accepts.
Related
The script step can run against a local model via Ollama instead of a paid API. For what else is climbing in the ecosystem, see the daily trending digest and the weekly report, or browse LLM tooling.
FAQ
Does it generate video footage with AI? No. It assembles copyright-free stock clips, an AI-written script, synthesized voice, subtitles, and music. The visuals are sourced, not generated.
What API keys do I need? At minimum a stock-footage key (Pexels) and a language-model provider key. A local model through Ollama can cover the script step.
Why did my run fail at the audio or compositing step? Those are the most-reported failure points. Check the TTS provider configuration first, then the video-combine step.
Is it free? The project is MIT-licensed. Your costs are the third-party model and footage services you plug in.