Automated short-form video, an early framework
ShortGPT is a framework for automating short-form video creation: it chains an LLM to write a script, text-to-speech for narration, captions, and stock footage into a finished vertical video. It was one of the earlier projects in this space, with engines for Shorts and TikTok content, longer videos, and video translation, and it supports many languages. At 7,405 stars as of 2026-06 it still draws attention as a reference design.
The headline you need before investing time: maintenance has stalled, which changes how you should approach it.
What it does
The pipeline is script to voice to captions to footage to composition. ShortGPT exposes engines, most notably a content-short engine that produces Shorts and Reels by generating a script with an LLM, narrating it with TTS, transcribing for captions, and assembling B-roll from a stock source. A translation engine re-voices existing videos in other languages. The design is coherent and worth reading as a blueprint for this kind of automation.
Install
ShortGPT is run via Docker, with API keys supplied through environment variables:
docker build -t short_gpt_docker:latest .
docker run -p 31415:31415 --env-file .env short_gpt_docker:latestYou configure keys for an LLM (OpenAI, with later Gemini support), optionally ElevenLabs for premium TTS (EdgeTTS is a free alternative), and a stock-footage source like Pexels. It also depends on ffmpeg and Whisper for transcription.
Maintenance has stalled
This is the part that should shape your decision. The last meaningful commits were in early 2025, and the tracker carries dozens of open issues, including Docker initialization failures, missing dependencies, and a broken YouTube audio download, none of which are being fixed. So while the architecture is sound, expect to debug your way to a working setup yourself, and do not count on upstream support. For a project whose whole value is automation that just runs, a stalled maintainer is a significant caveat.
The risks worth naming
Two risks deserve explicit mention. First, cost: you pay per video for the LLM and, if you use it, premium TTS, plus the time of debugging a stalled stack. The free path (EdgeTTS plus a free LLM tier) works but caps quality. Second, and more important, platform policy: YouTube and TikTok restrict mass automated and undisclosed AI-generated uploads, and bulk posting can trigger throttling or bans. Stock footage from Pexels is generally free to use, but you remain responsible for the rights of any generated content and for disclosing AI generation where required. Treat this as a tool for personal, low-volume, reviewed output, not an unattended content farm.
Where it fits, and where it doesn’t
ShortGPT is worth your time as a learning reference for how an automated video pipeline is structured, or for a developer comfortable debugging a Dockerized Python stack who wants to experiment.
For anyone who wants something that reliably works today, the stall is disqualifying: the Docker setup is fragile, issues go unanswered, and the dependencies have drifted. A maintained alternative is the safer choice, which is exactly what the comparison below points to.
ShortGPT versus the active alternatives
| ShortGPT | MoneyPrinterTurbo | MoneyPrinter | |
|---|---|---|---|
| Stars | 7,405 | 87,305 | 13,530 |
| Maintenance | stalled, last commit early 2025 | very active | maintenance mode |
| Setup | Docker, fragile | one-click, Docker, Colab | local, Ollama-first |
| Models | OpenAI, later Gemini, EdgeTTS | many (OpenAI, Gemini, DeepSeek, Ollama, and more) | Ollama-first |
| Best for | reading the architecture | actually generating videos | local, privacy-first runs |
Counts are from GitHub as of June 2026. MoneyPrinterTurbo is the far larger, actively maintained choice for actually generating short videos today. MoneyPrinter is a lighter, local, Ollama-first option. ShortGPT’s value now is as a reference, not a tool you should depend on.
The star curve
The star curve rose during the AI-content boom when automated short-video generation was a hot idea, then flattened as maintenance stopped. The gap between the star count and the current usability is the thing to keep in mind: stars reflect past interest, not present health.
What the issue tracker warns you about
The open issues map the pain directly. Docker initialization fails and a required image dependency is missing, so a clean setup is not guaranteed. The YouTube audio download is broken and the metadata step can hang. macOS installs have stalled for very long times. And general automation runs fail with type errors in the current state. Because the project is not being maintained, these are unlikely to be fixed, so budget for self-debugging or pick a maintained tool.
Related
For a maintained, popular alternative, see MoneyPrinterTurbo. For a lighter, local-first option, see MoneyPrinter. For programmatic, code-defined video instead, see Remotion. For what is trending, see the daily digest and the weekly report.
FAQ
Does ShortGPT still work in 2026? Maybe, but with effort. Maintenance stalled in early 2025, and open issues include Docker failures and a broken YouTube audio download. You can likely get it running by debugging, but there is no upstream support, and a maintained alternative is more reliable.
ShortGPT vs MoneyPrinterTurbo, which should I use? MoneyPrinterTurbo, if you want something that works today. It is far larger, actively maintained, supports many models, and is easier to set up. ShortGPT is best read as an architectural reference now.
How much does ShortGPT cost to run? You pay per video for the LLM and, optionally, ElevenLabs TTS; the free path uses EdgeTTS and a free LLM tier at lower quality. Stock footage from Pexels is free. The bigger hidden cost is the time to debug a stalled stack.
Is it legal to mass-produce videos with ShortGPT? Generating them is fine, but uploading matters. YouTube and TikTok restrict mass automated and undisclosed AI uploads, which can trigger throttling or bans. Keep volume low, disclose AI generation where required, and respect each platform’s terms.
What does ShortGPT need to run? Docker, an LLM API key (OpenAI, with later Gemini support), optionally an ElevenLabs key for premium TTS, a stock-footage key like Pexels, plus ffmpeg and Whisper. Expect to troubleshoot dependencies given the stalled maintenance.