To deinterlace a video, run it through a deinterlacing filter that rebuilds full progressive frames from its interlaced fields, then encode the result. In FFmpeg, a good default is -vf bwdif. It outputs one frame per field (for example 50 fps from 25i), which keeps the smooth motion of the original. Use yadif as a widely supported alternative, and check first whether the video is actually interlaced or telecined film, because the two need different treatment.
Interlacing is a leftover from analogue TV, but it is still everywhere in DVDs, camcorder tapes, broadcast recordings and some cameras. This guide explains how to recognise it, how to choose between deinterlacing methods and output frame rates, and how to deal with telecined film.
What interlacing is
An interlaced frame contains two fields captured at different moments: one made of the odd lines and one of the even lines. A 25i PAL signal, for example, carries 50 fields per second, so motion is sampled 50 times a second even though there are only 25 frames. CRT televisions showed the fields one after another, and the result looked smooth.
Modern screens are progressive and show whole frames at once. If an interlaced frame is shown as-is, the two fields from different moments are woven together, so anything moving gets jagged horizontal lines, known as combing. Deinterlacing reconstructs proper progressive frames, so the video looks right on modern displays and compresses efficiently. Interlaced combing wastes a lot of bitrate.
Step 1: confirm the video is interlaced
Common interlaced sources include DVDs (576i or 480i), DV and HDV camcorder tapes, broadcast TV recordings (including 1080i HD), digitised VHS, and MPEG-TS or M2TS files from some cameras and set-top boxes. To check:
- Look at motion. Pause on a fast-moving scene. Comb-like lines on moving edges are a sure sign.
- Check the metadata.
ffprobe -v error -select_streams v:0 -show_entries stream=field_order input.mkvreportsttorbb(ortb/bt) for interlaced andprogressivefor progressive. Flags can be wrong, so don't rely on them alone. - Analyse the frames. FFmpeg's
idetfilter counts frames that look interlaced:ffmpeg -i input.mkv -vf idet -frames:v 1000 -an -f null -. The "Multi frame detection" line shows counts of top-field-first, bottom-field-first and progressive frames.
idet reports mostly progressive frames, leave the video alone, or use a deinterlacer that only processes frames flagged or detected as interlaced.Step 2: choose the output frame rate
Deinterlacers can output in two ways:
- One frame per field (double rate) - 25i becomes 50p and 29.97i becomes 59.94p. Every moment of motion is kept, so video-style content such as sport, live TV and home video stays as smooth as it was on a TV. This is the most faithful option.
- One frame per frame (single rate) - 25i becomes 25p. The file is smaller and plays on more devices, but half the temporal information is discarded, so motion looks less fluid.
For home video and broadcast content, double rate usually looks best. Single rate is a reasonable compromise when file size or compatibility matters more.
Step 3: pick a deinterlacer
| Filter | Approach | Notes |
|---|---|---|
bwdif | Motion-adaptive, based on yadif and w3fdif | Good quality and speed; a sensible default. Outputs one frame per field by default. |
yadif | Motion-adaptive, spatial and temporal checks | Long-standing and widely used; outputs one frame per frame by default (yadif=1 for double rate). |
w3fdif | Weston 3-field, from BBC research | Double rate by default. |
estdif | Edge slope tracing | Spatial-only; interesting for some content. |
nnedi | Neural-network edge interpolation | Can give clean edges but is slow and needs a separate weights file. |
| QTGMC (AviSynth/VapourSynth) | Motion-compensated, multi-stage | Often considered the best quality available, but very slow and outside FFmpeg itself. |
A good-quality double-rate encode with FFmpeg:
ffmpeg -i input.mkv -vf bwdif=mode=send_field -c:v libx264 -crf 18 -preset slow -c:a copy output.mp4
Single rate, if you need 25p or 29.97p:
ffmpeg -i input.mkv -vf bwdif=mode=send_frame -c:v libx264 -crf 18 -c:a copy output.mp4
The audio doesn't change length, so it can usually be copied unless the container needs a different codec.
Field order matters
Interlaced video is either top field first (TFF) or bottom field first (BFF). DV is typically BFF, while most HD broadcast and DVD content is TFF. Deinterlacers normally read the order from the stream, but if the flag is wrong the output will judder back and forth, with motion stepping forwards and backwards. If you see that, override it, for example -vf "setfield=tff,bwdif" or -vf "setfield=bff,bwdif", and compare.
Telecined film is different
Films and many TV dramas shot at 24 fps were transferred to 29.97i NTSC using 3:2 pulldown (telecine), which spreads 4 film frames across 10 fields. The result shows combing on some frames but not others. Deinterlacing it works, but it throws away detail and leaves juddery motion. The right fix is inverse telecine: match the fields back into the original progressive frames and drop the duplicates, giving 23.976p:
ffmpeg -i input.vob -vf "fieldmatch,yadif=deint=interlaced,decimate" -c:v libx264 -crf 18 -c:a copy output.mkv
Here fieldmatch rebuilds the frames, yadif=deint=interlaced cleans up any frames that couldn't be matched, and decimate removes the one duplicate in every five frames. PAL film releases are usually simpler: the film was sped up to 25 fps and the fields often pair into progressive frames already. Our guide to changing video frame rate covers related conversions.
Order of operations
- Deinterlace before scaling. Resizing interlaced video mixes the two fields together and makes clean deinterlacing impossible afterwards.
- Crop and denoise after deinterlacing. Most denoisers and sharpeners expect progressive frames.
- Use enough bitrate. Older SD sources often carry noise. A lower CRF value or a light denoise helps the encode look clean.
- Keep the original. Deinterlacing is irreversible. For valuable tapes or broadcasts, consider keeping a lossless copy of the interlaced capture as well. Our guide to FFV1 lossless video archiving explains how.
Should you keep it interlaced instead?
For broadcast delivery that requires interlaced formats, or for archive masters, keeping the original interlaced stream (and encoding it as interlaced if you must re-encode) can be the right choice. For anything watched on computers, phones, smart TVs or the web, deinterlacing to progressive is the normal and recommended approach. Many devices deinterlace on playback, but quality varies a lot, and web platforms expect progressive uploads.
Doing it with Movie Encoder
Movie Encoder's media info shows each file's codec, frame rate and stream details, and deinterlacing is one of its built-in filters, along with crop, resize, denoise and frame rate conversion. The before/after split preview and short sample encodes let you check a fast-moving scene for combing or judder before you run the whole file.
The command preview shows the exact FFmpeg filter chain, so you can confirm the order of operations. Your original capture is never overwritten. For tape archives, Archive & Verify can create lossless FFV1 copies in MKV with SHA-256 checksums.
Frequently asked questions
How do I know if my video is interlaced?
Pause on fast motion and look for comb-like horizontal lines on moving edges. FFmpeg's idet filter can also analyse the frames and report whether they look interlaced or progressive.
Should I deinterlace to 25 or 50 fps?
Outputting one frame per field, giving 50 fps from 25i, keeps the original smooth motion and suits home video and sport. 25 fps gives smaller, more widely compatible files but less fluid motion.
What is the difference between deinterlacing and inverse telecine?
Deinterlacing rebuilds frames from fields captured at different moments. Inverse telecine reassembles film frames that were spread across fields by 3:2 pulldown, recovering the original 23.976 fps progressive video.
