Glue code that turns text prompts into TikTok-ready clips

What it does

This is a Python pipeline that generates short-form videos from a text prompt. It delegates every step: an LLM writes the script, a TTS service narrates it, Pexels supplies B-roll, and a speech-to-text engine burns in captions. The result is a rendered MP4 tuned for YouTube Shorts, Reels, or TikTok.

The interesting bit

There is no custom video model here—just pragmatic glue code that knows which APIs to call and in what order. The README is candid about its commercial sibling, pitching a paid Vadoo API as the “recommended” path, but the open-source route still works via Google Colab if you bring your own keys.

Key highlights

• Swappable providers: choose between OpenAI, Groq, or Gemini for scripts; EdgeTTS or ElevenLabs for voice; Whisper or Deepgram for captions.
• Free-tier viable: EdgeTTS and Whisper cost nothing, though Pexels and an LLM key are still mandatory.
• Automatic B-roll: fetches relevant stock footage from Pexels to match the generated script.
• Aspect ratio presets: outputs 9:16 portrait for mobile feeds or 16:9 landscape for standard video.
• Browser-ready: includes a Colab notebook so you can avoid local dependency wrangling.

Caveats

• Not self-contained: you must supply API keys for Pexels and at least one LLM provider to get a working pipeline.
• Cross-platform friction: Windows users need a separate installation guide, implying Linux and macOS are the smoother paths.
• The README prioritizes the Premium API upsell, which can make the open-source setup feel like an afterthought.

Verdict

Worth a look if you want to prototype faceless short-form content without touching video editing software, or if you need a scriptable assembly line for social clips. Avoid if you were hoping for a fully local, weights-included diffusion model—this is orchestration, not generation.