🎨 Colorization & Enhancement Methods
At Public Domain Colorizer, we strive to breathe new life into historical public domain films by applying advanced artificial intelligence (AI) techniques to colorize, enhance, and restore old black-and-white videos.
This page outlines the full process we use, the tools involved, and the philosophy behind our work.
🔧 Tools & Technologies We Use
Our workflow is based on open-source and AI-powered tools, run mostly on cloud services like Google Colab. The key components include:
1. DeOldify – Colorization
-
We use DeOldify, an open-source deep learning model built on PyTorch.
-
It intelligently predicts natural-looking colors based on context, textures, and historical data.
-
It produces realistic skin tones, skies, fabrics, and objects without manual painting.
2. EDVR – Video Restoration
-
EDVR (Enhanced Deformable Video Restoration) is used to remove flickering, artifacts, and stabilize frames.
-
It improves temporal consistency, making motion more fluid in the final video.
3. ESRGAN – Upscaling
-
We upscale videos using ESRGAN (Enhanced Super-Resolution Generative Adversarial Networks).
-
This adds clarity, sharper edges, and finer details — bringing old films closer to modern visual quality.
4. Google Colab – Processing Platform
-
All processing is done using Google Colab notebooks.
-
It allows us to access powerful GPUs and automate parts of the enhancement pipeline for free.
🖼️ Step-by-Step Colorization Pipeline
-
Video Sourcing: We download verified public domain content from trusted archives like Archive.org.
-
Frame Extraction: Videos are split into frames using tools like FFmpeg.
-
Colorization: Frames are processed through DeOldify (Artistic or Stable model) to generate color.
-
Restoration: Colorized frames are enhanced using EDVR to reduce flicker and noise.
-
Upscaling: ESRGAN is applied to upscale frames to 720p or 1080p resolution.
-
Reassembly: All enhanced frames are recompiled into a video with optional silent music.
-
Metadata Addition: We add contextual metadata, alternative titles, and historical info to the final post.
-
Publishing: The completed video is uploaded to platforms (like YouTube or Archive.org) and embedded on our blog.
🧠 Why We Do This
-
🕰️ Preservation: These films are cultural treasures that deserve renewed visibility.
-
🎓 Education: Colorized versions help modern audiences understand history better.
-
💡 Innovation: We embrace AI not just to entertain, but to extend the lifespan of forgotten art.
📬 Contact
If you believe a film may not be in the public domain or you'd like to collaborate, reach us at:
0 Comments