Build Custom AI Video Generator -- 2

I want a self-contained solution that can take a variety of inputs—plain text scripts, voice recordings, or even existing clips—and automatically turn them into polished videos. The finished system should let me pick a style on the fly (fully animated scenes, narrated visuals, or simple text-on-screen with background music) so I can produce educational, marketing, or entertainment pieces without touching a video editor.

Here is what I have in mind:
• A web-based or desktop interface where I can paste a script, upload audio, or drop a video file.
• Server-side or local processing that handles speech-to-text, scene generation, voice-over synthesis, background music selection, and final rendering up to 1080p.
• Modular architecture so I can swap models later—e.g., replace the TTS engine, upgrade the animation model, or add new templates—without rewriting the whole app.
• A small set of starter templates for each style that I can tweak (logo, brand colors, font, outro).
• Output in MP4 and MOV with subtitle files.

Acceptance criteria
1. I feed a short script or voice note and receive a rendered video in one of the three styles.
2. A README explains how to install, configure, and extend the models or assets used.
3. All third-party libraries are listed with licenses and version numbers.

Feel free to recommend specific AI stacks—PyTorch, TensorFlow, ffmpeg, Stable Video, ElevenLabs, etc.—or any SaaS APIs if they speed things up. The key is that I end up owning the deployable codebase and can run it on my own hardware or cloud account.