Colorize Old Black-and-White Video

AI-driven black-and-white to color video conversion that runs right in your browser. The neural network paints natural skin tones, sky, foliage, and fabrics — pair it with film-stock presets to match the era of the original footage.

How to colorize a black-and-white video

1

Upload your black-and-white video — MP4, WebM, MOV, AVI, MKV, or M4V

2

Pick a quality preset (Fast, Standard, or High) and an optional film-stock look

3

Check the colors on the live preview or run "Preview 5 seconds"

4

Tweak skin tone or drop a reference image to bias colors, then export

5

Download as MP4, WebM, GIF, or a ZIP of PNG frames

Restore color to vintage footage with a neural network

Drop video file here
MP4, WEBM, MOV, MKV, AVI, M4V, 3GP, OGV ·
Try with a sample clip
Don't have a B&W video handy? Click any clip below — it loads on demand, the page itself stays light.
AI model not yet available The colorization model is being prepared. Check back soon, or follow the project for updates.
Original Colorized
Preparing preview...
Show original video
Scenes
Click a thumbnail to preview colorization at that moment.
Content type
Best for photographs and live-action film.
Film stock look
Color & tone
Intensity
100%
Saturation
+15
Boosts color intensity. AI colors can look muted on real footage; right for vivid, left for natural.
Skin tone correction
0
Nudges face/hand skin toward natural warmth. Drag right for warmer, left for cooler.
Reference palette (optional)
Drop a color photo to guide the palette
Recolour or mute a color the AI got wrong — e.g. a brown sky.
Color to fix
Range
New color
Strength
AI processing
Stabilises color across frames so it does not shimmer or drift between shots. Recommended for video.
Stabilisation strength
60%
Pre-equalizes dark or washed-out footage before AI inference for cleaner colors.
Auto-detects people; the background fades to grayscale.
Background fade
100%
Paint colored strokes on the first frame; the AI uses them as guidance for the whole video.
Film & effects
Adds analog noise to mimic film stock — pairs well with vintage presets.
Grain intensity
50%
Common on VHS rips and old broadcast recordings.
Export
Quality
Every frame — the most accurate color.
Output format
Resolution
My saved looks
No saved looks yet — tweak settings and save your favorite combos.

Paint color hints on the keyframe

Each painted area biases the AI output toward that hue. The hints apply to every frame at the same location — works best for static-camera footage.
Initializing...

Colorized result

This is a preview — only the first seconds are colorized. Happy with the colors? Start the full render.

Per-frame correction

1 / 1
Paint over individual frames to override the AI color. The change applies to that single frame only.
Don't know what to try?
750+ tools — open a surprise
🎲 Surprise me

What this colorizer does

Neural network runs right in your browser
The model downloads once (~108 MB) and is cached for offline reuse; devices with hardware acceleration process noticeably faster.
6 film-stock presets: Kodachrome '55, Technicolor '60s, Faded '70s, Cinema noir, Sepia warm, Modern
Layer a period-accurate color grade over the AI result and dial in its intensity to match the era of your footage, with optional film-grain overlay.
Skin-tone correction slider for natural-looking faces
Nudge faces and hands warmer or cooler to fix the washed-out "zombie face" that automatic colorization often gives to skin.
Reference image palette — drop a colored photo to bias the output
Feed in a color photo at least 64×64 pixels and the tool extracts its palette to steer the colors of the whole clip.
Multiple export options including video, animation, and frame archives
Save the result as MP4 (H.264), WebM (VP9), an animated GIF for short clips, or a ZIP of PNG frames, with the original audio copied across.
Before/after wipe slider on the result preview
Drag a split slider across the result, or switch to side-by-side, to compare the original black-and-white against the colorized version.

Why use this colorizer

AI video colorizers typically run behind a queue and watermark the output. This one runs right in your browser: the model loads once and works offline thereafter. No queue, no watermark. Pair it with film-stock LUTs to match the era of your footage and a skin-tone slider to fix the typical AI-colorization "zombie face" look.

Frequently Asked Questions

Why does colourising a video cost so much more than a photo?

Because the network has no idea what a video is: it colors one still image at a time, and a minute at 30 frames a second is 1800 separate jobs. That also explains the flaw you have to watch for — the model decides a coat is brown in one frame and gray two frames later, and the eye reads that flicker instantly even though each individual frame looks fine. Short clips of a stable scene come out far better than long ones with cuts and camera moves. The model itself is a download of about a hundred megabytes that happens once and is then kept, and the work runs on your own machine, so a laptop with hardware acceleration will be several times faster than one without.

What makes a clip too long for colourising?

Every frame goes through the network, so the cost is duration times resolution and nothing else — a thirty-second home movie at 480p is comfortable, a ten-minute feature clip at 1080p is not. The model itself downloads once, about 108 MB, and is cached afterwards, so the wait you notice on a second clip is the painting, not the loading. Trim to the passage worth coloring before you start.

Is the original audio preserved?

Yes — the audio track is copied unchanged into the MP4 and WebM outputs. GIF and the PNG-frame ZIP have no audio container, so audio is dropped automatically for those formats.

The faces came out gray and lifeless. Can that be fixed?

That is the usual first result, and there are two dials for it. The skin-tone slider nudges faces and hands warmer or cooler until they stop looking like a waxwork — a small move is normally enough. If the whole picture is off rather than just the people, drop in a color photo from the same era as a reference and the palette is biased towards it; anything from 64×64 pixels up works.

💡 Want us to improve this tool just for you?

We can! Just send us a quick message with your idea. If you'd like to discuss it in detail, leave your email and we'll get back to you. You can stay anonymous.

How do you rate this tool?

3.7 (3 ratings)
Thank you for your rating!
Want to share more? Leave a comment!
Thank you! Your comment will appear after moderation.
Published Updated Author: