How to Colorize Black-and-White Video with AI (Step by Step)

To colorize black-and-white video with AI, clean the footage first (deinterlace, remove grain, steady the shot), then run an AI colorizer and judge the result on faces, skies, and foliage before you export. AI colorization predicts likely colors from what's in the frame.

It can't know the real color of a dress, a car, or a uniform, so treat the result as an interpretation, label it as colorized, and keep the original.

• AI colorization predicts plausible colors from shapes, textures, and context. The result is an interpretation, not a record of the original colors.
• Clean the footage before you colorize it: deinterlace, remove grain, and steady jittery film first.
• Start at the default strength, then tune shot by shot while you watch faces, skies, and foliage.
• Label colorized versions and keep the black-and-white original.
• Use footage you own, have permission to use, or that is in the public domain.

AI video colorization is the process of adding plausible color to black-and-white footage with a model trained on large numbers of color images.

The model looks at each region of a frame (its brightness, texture, edges, and the objects around it) and predicts the colors that usually appear in similar scenes: skin tones for faces, blue for open sky, green for leaves. For video, the extra challenge is consistency. A coat that turns from brown to blue between two frames is more distracting than a slightly wrong color that stays put.

A black-and-white frame records brightness, not hue. A red scarf and a green scarf of the same brightness look identical in gray. That leads to a simple rule:

• Things with typical colors usually come out believable: skin, sky, grass, trees, water, wood, brick.
• Things whose color was a choice come out as guesses: clothing, cars, painted walls, flags, team jerseys, signage.

If the true color matters (a family member's wedding dress, a regiment's uniform, a brand's logo), plan to check it against a reference photo or record and correct it by hand.

Colorizers give steadier results on clean, stable, progressive frames. Do these first:

1. Deinterlace. Tape and broadcast transfers are often interlaced. In KwaFlux, deinterlacing runs automatically as the first step when the source is interlaced.
2. Remove grain and dust. Heavy grain gives the colorizer noisy edges to follow. In KwaFlux, the Old Film card runs Film Grain Removal. Because the scene type is one choice per pass, run grain removal as its own pass before colorizing.
3. Steady jittery film. Gate weave and handheld shake make colors harder to keep consistent. KwaFlux's Video Stabilization (under AI Toolbox › Smart Effects) previews 5, 15, or 30 seconds before you commit.
4. Check for tints. Sepia or tinted film isn't truly black and white. If a preview comes out muddy, remove the tint in an editor first, then colorize the neutral grayscale version.
5. Use footage you have rights to. Your own family films, footage you've licensed, or material in the public domain. Copyright depends on the country and the work, so check before you publish.

Pro tip: After each pass, use Continue with to send the result to the next tool. The output is added as a new version under the source clip, and the original stays untouched.

1. Open the Colorize B&W card. In the sidebar, go to Quality Enhancement and click Colorize B&W.
2. Import your clip. Click Add video or drag it in. The card opens Quality Repair with Repair Scene set to Scene Repair and Scene Type set to Colorize.
3. Leave Colorization Strength at the default for the first test. The slider runs from 0 to 1 and starts at 0.6. Higher values give richer color.
4. Preview three kinds of shots. Render Preview 1s, 3s, or 5s on: a close-up face, to judge skin; a wide shot with sky or trees, to judge large areas; a shot with movement, to check that colors hold steady. Compare side by side, or drag the slider between before and after.
5. Adjust and preview again. Lower the strength if skin looks orange or colors look painted on; raise it if large areas stay gray or brown.

For video, KwaFlux colorizes with a path that reads neighboring frames together, which is meant to keep colors steadier from frame to frame. Still check moving shots in the preview, because consistency varies with the footage.

Sidenote: Photos work too. Import a black-and-white image into the same workspace; for images, the preview shows the whole picture. See the image enhancer for more on stills.

One setting rarely suits a whole reel. A bright beach scene and a dim interior respond differently.

• Group shots by look: exteriors, interiors, close-ups.
• Export each group with its own strength. Use Export Range to set a start and end for each section, or cut the reel into scenes in your editor first.
• Watch the cuts. When two shots of the same place sit side by side, make sure the same wall or coat doesn't change color across the cut.

1. Set export options. In Export Settings, choose the folder, container, and codec. If you'll upload to YouTube, its guidance recommends BT.709 as the color space for SDR uploads.
2. Label the result. Put "colorized" in the file name, video title, or description, especially for family archives and historical material.
3. Keep the black-and-white original. It's the real record, and future tools may do a better job from it.
4. Optional: add detail afterward. In KwaFlux 1.0.8, Scene Repair (except Anime) keeps the source resolution. If you want a sharper or larger picture, send the colorized version to Smart Repair with Continue with, then upscale.

When a specific color has to be right, correct it by hand in a color-grading tool. Blackmagic lists luma, HSL, and 3D keyers among the color tools in the free DaVinci Resolve, and puts Magic Mask object isolation in the paid Studio version.

A practical hybrid workflow:

1. Colorize the whole clip with AI.
2. In your editor, isolate the object that's wrong (a uniform, a car, a sign) with a qualifier or a window.
3. Track the selection through the shot and shift its hue to the reference color.
4. Leave everything else as the AI produced it.

This keeps hours of manual work to the few objects that matter.

• It can't recover the original colors. It predicts likely ones.
• It can't add detail. Soft footage stays soft; repair or upscale separately.
• It can't make footage historically accurate. Treat colorized archives as interpretations and say so.
• It can't keep one object's color consistent across separate shots on its own. Check cuts and correct where needed.

Good colorization is mostly preparation and judgment: clean frames, a sensible strength, careful checks on faces and cuts, and an honest label. Give the AI the boring 95% of the work and spend your own time on the few colors that matter.

KwaFlux restores old video on your own computer, from deinterlacing and grain removal to colorization and upscaling. AI processing runs locally on your GPU — media files are not uploaded for processing. It runs on Windows 10 (64-bit, version 1809 or later), Windows 11, and macOS 12 or newer on Apple Silicon. Previews need no account: 1, 3, or 5 seconds for video, and the whole picture for images. Signing in with an active paid plan is required only to export, convert, repair a file, or download a file from a link.

Related reading: restore old footage with AI and the old home video workflow.