ScriptToScreen — Consistent Character AI Video Pipeline
Link to open source: https://github.com/DollyQx/Vibe-Video-Flow
Link to Live Project: https://vibe-video-flow-98rj.vercel.app/
The Vision: In the world of AI video, "character drift" is the biggest hurdle. ScriptToScreen is an automated pipeline designed to take a raw text script and turn it into a cinematic video where the protagonist remains visually identical from the first frame to the last.
Why we built it: Current AI models generate great 10-15 second clips, but they fail to maintain consistency over a longer narrative. We built this to bridge the gap between storytelling and high-fidelity video production, automating the decomposition, generation, and character-locking process.
How it works (The Tech):
-
Narrative Orchestration: We use OpenAI (GPT-4o) to intelligently parse long scripts into structured visual scenes.
-
Character Lock Architecture: Using the Pixazo API, we generate a master reference image of the character (Dolly) and anchor every subsequent video generation call to this reference.
-
Hybrid Deployment: Built with Next.js on the frontend (Vercel) and a robust Node.js backend on Render to manage heavy API tasks and bypass serverless timeout limits.
Key Features:
-
Zero-drift character consistency.
-
Automated script-to-scene decomposition.
-
Seamless sequential video playback.
-
High-speed cloud-based processing.
