RE Colorizer

RE Colorizer is a Windows application that automatically adds color to black-and-white photographs and videos using artificial intelligence. All processing occurs entirely on the user's computer without sending files to external servers. The software supports JPG, PNG, MP4, AVI, and MOV formats. It can process single images, entire batches of photos, and videos of any length. The underlying technology uses neural networks trained on millions of images to assign realistic colors to objects, faces, landscapes, and textures.

★★★★★

4.0(2 ratings)

Download RE Colorizer (Official links)

File size: 1114 MB

The latest version of RE Colorizer is: 2.5.1

Operating system: Windows

Languages: English

Developer: Recolorizer.com

Price: $39.00 USD

Frame-by-frame video colorization. This function extracts each frame from a video file, applies the colorization model individually, and then reconstructs the video while preserving the original audio track. The process ensures color consistency between consecutive frames, avoiding flickering or abrupt tonal shifts. The user selects a video file, and the program generates a new colorized MP4 file with the same duration and base resolution.
Batch image processing. Allows selecting an entire folder containing multiple black-and-white photographs and applying colorization to all of them sequentially without manual intervention. Each image processes with the same configured parameters (model, render factor, color stabilization). Upon completion, all colorized images save to a designated subfolder, preserving original filenames and file structure.
Adjustable render factor. Controls the internal resolution at which the neural network processes each image. Values between 10 and 20 produce fast colorizations with lower resource consumption. Values between 21 and 35 offer a balance between speed and fine detail in edges and textures. Values between 36 and 45 maximize color quality in faces, eyes, and hair, though they increase processing time and required video memory.
Stable model for vintage photography. This model prioritizes color consistency in images with high noise levels, grain, stains, or physical deterioration. It reduces saturation in ambiguous areas to avoid artificial colors or incorrect tone patches. It is optimized for scanned photographs from the early 20th century, studio portraits with neutral backgrounds, and landscapes with overexposed or underexposed skies.
Artistic model for maximum vibrancy. Generates more saturated and contrasted colors, suitable for originally creative or artistic material. This model tends to produce warmer skin tones, more intense greens in vegetation, and deeper blues in skies and water. It may introduce color variations not present in reality but visually appealing. Recommended for illustrations, concept art, or author photography where historical fidelity is not the primary concern.
Frame-to-frame color stabilization for video. During video colorization, this function analyzes adjacent frames and adjusts differences in hue, brightness, and saturation to avoid perceptible fluctuations. It uses a temporal filter that smooths abrupt color transitions caused by lighting changes or fast motion. The result is a video where stationary objects maintain the same color over time, improving the perception of naturalness.
Pre-colorization noise reduction. Before applying the AI model, the software executes a digital noise and grain removal filter on the original image. This step improves neural network accuracy by eliminating artifacts that could be misinterpreted as edges or textures. Particularly useful for photographs taken at high ISO sensitivity or videos captured in low-light conditions. The noise reduction level applies automatically based on image analysis.
Post-colorization edge and texture enhancement. After assigning colors, the software applies a selective sharpening filter that reinforces object edges without amplifying noise. This function restores fine details such as wrinkles in clothing, facial hair, tree leaves, or wall textures that might have been smoothed during the colorization process. The result maintains the original image clarity while adding color.
EXIF metadata preservation in images. When saving a colorized photograph, the program keeps all original file metadata intact, including capture date, camera model, exposure settings, original white balance, and GPS coordinates if present. This function is critical for photographers, archivists, and historians who require traceability and preservation of technical information alongside the colorized version.
GPU processing with NVIDIA CUDA. When the system has a CUDA-compatible NVIDIA graphics card with at least 6 GB of video memory, the software executes all AI models directly on the GPU. This reduces the colorization time for a high-resolution image from several minutes to less than 5 seconds. For videos, GPU acceleration allows processing one minute of footage in approximately 30-60 seconds, depending on resolution and configured render factor.
Automatic CPU mode without GPU. On computers without a dedicated graphics card or with insufficient GPU memory, the software automatically switches to CPU processing. It uses AVX2 instructions and multi-core optimizations to execute the same colorization models, though with longer processing times. A 1920x1080 image may take between 2 and 5 minutes on CPU. CPU mode allows using RE Colorizer on older laptops, workstations without a GPU, or machines with non-compatible graphics cards.

RE Colorizer was originally created by developer Philippe Mathis under PhM 1000 Software Engineering. Development began in 2019 as a project to bring DeOldify technology to users without programming knowledge. The first stable Windows version was released in 2020. The program is written primarily in Python for neural network inference models, using the fastai and PyTorch libraries. The graphical user interface and Windows integration logic were developed in C++ and C#. The PRO version added support for CUDA acceleration and full video processing starting in 2021.

Alternatives to RE Colorizer:

Pinocast — Free Download. Neural makeup for live video

RE Colorizer

Alternatives to RE Colorizer:

Pinocast

ON1 Photo RAW

Topaz Photo

VideoProc Vlogger

AliveColors

Photo Toolbox

ACDSee Photo Studio

SkinFiner