M2TS to MJPEG Converter

Every Frame of Your AVCHD Clip as a Standalone JPEG

An .mjpeg file is not a container in any meaningful sense — it is a run of complete JPEG images written end to end with nothing between them. Convert a four-second 25 fps M2TS and you get exactly 100 JPEGs in one file; we counted the FF D8 FF start-of-image markers to check, and cutting the file at any marker yields a valid .jpg. Nothing needs to understand the file's structure to take it apart, because there is no structure.

That is the whole reason to choose this target over MP4 or MOV. Frame extraction for a machine-learning dataset, photogrammetry input, motion analysis, contact sheets, forensic review — anything that wants "give me the pictures, one per frame, decodable in any order" is what MJPEG is for. It is a terrible way to watch video and an excellent way to process it.

What Crosses Over from an M2TS, and What Does Not

M2TS is the Blu-ray and AVCHD stream format: a transport stream with a BDAV twist, where every 188-byte packet carries a 4-byte arrival-timestamp prefix, so the sync byte sits at offset 4 rather than at the start. We checked a test file and the first bytes read 0e bf 46 22 47 — exactly that pattern. Inside is normally H.264 video with AC-3 or LPCM audio at 48 kHz, from a camcorder or a Blu-ray rip.

The audio loss is the one that catches people out. It is not a setting: a raw Motion JPEG stream is JPEG images and nothing else. The page does render an Audio Codec dropdown — it is shared across every video target — but nothing you select there reaches the file. If you want the AVCHD soundtrack alongside intraframe JPEG frames, run M2TS to MOV and set the Video Codec to MJPEG; a real container holds the audio and the timing.

How to Convert M2TS to MJPEG

1. Upload Your M2TS File: Drag the file onto the page or click "+ Add Files". Camcorder clips are often split into several M2TS files by the recorder; each is converted independently, so queue them together if you want the same settings applied to all.
2. Switch File Compression to Constant Quality: MJPEG is driven by a JPEG quantizer, not a bitrate, and Constant Quality is what exposes that control — a slider running 2 to 31 and opening at 5. Leaving File Compression on Quality Preset leaves the quantizer at its default value instead.
3. Scale Down Under Video Resolution: This is the size lever that matters most on HD footage. Video resolution offers Keep original, Resolution Percentage, Preset Resolutions and an exact Width x Height; halving both dimensions cuts the per-frame data to roughly a quarter.
4. Trim and Convert: Trim starts at "Unchanged" and can export a single Start time / Duration range — with MJPEG, converting only the seconds you actually need is usually the difference between a workable file and an unworkable one. Click Convert and download the .mjpeg. No sign-up, no watermark.

Sizing the Job Before You Start

MJPEG has no compression between frames. A locked-off tripod shot where nothing moves still stores the entire scene again on every single frame, where H.264 would record "same as before" in a handful of bytes. That trade is the point — perfect frame independence — but on HD footage the arithmetic gets uncomfortable fast. Figures below are measured from a 1920 × 1080 25 fps test clip.

Real footage varies with how much detail is in the frame, but the shape holds: an hour of 1080p MJPEG is measured in tens of gigabytes whatever you do to the slider. Note also that the slider's own 2–31 range is remapped onto JPEG's quantizer scale rather than used directly, so neither the finest quantizer (2) nor the coarsest (31) is reachable from the page. Your three real levers are the slider, the frame size, and the trim.

Interlaced Footage Comes Through Interlaced

A great many M2TS files — particularly from consumer AVCHD camcorders set to 1080 60i or 1080 50i — are interlaced. There is no deinterlacing filter anywhere in this pipeline, so each output JPEG holds both fields woven together exactly as the camera recorded them, and any horizontal movement shows the classic comb pattern on its edges.

For a frame-extraction workflow this can be exactly what you want: you are getting the sensor data as recorded rather than as an algorithm reconstructed it, and some analysis pipelines prefer to do their own field separation. For anything that will be looked at, it is not — the combing is baked into every JPEG and no later step can cleanly undo it. If your source is interlaced and the frames are for human eyes, deinterlace locally first and upload the progressive result.

Frequently Asked Questions

How do I turn the .mjpeg back into numbered image files?

Any tool that can scan for JPEG boundaries will do it, because each frame starts with FF D8 FF and ends with FF D9. One ffmpeg command handles it — ffmpeg -i clip.mjpeg frame%05d.jpg — and so does a short script that splits on the markers. Nothing else about the file has to be understood, which is precisely why frame-extraction workflows like this format: there is no index to parse, no container to seek through, and no codec state that carries between frames.

Why does my MJPEG play at the wrong speed?

Because a raw Motion JPEG stream stores no timestamps. There is no header recording that the camcorder shot at 59.94 or 25 fps, so a player or a tool like ffprobe assumes a default — commonly 25 fps — regardless of the source. A 59.94 fps clip written to .mjpeg and read back will play at less than half speed. The frame count is exact; only the timing is missing. If playback speed matters, use a container instead of the raw stream.

Is the MJPEG bigger than the M2TS I started with?

Substantially, essentially always. AVCHD's H.264 uses interframe prediction and long GOPs, so a camcorder clip of a mostly static scene is remarkably compact; MJPEG throws that away and stores the whole scene on every frame. In our 640 × 480 test the M2TS was 676 KB and the default-quality MJPEG was 2.85 MB from the same footage — more than four times larger — and that ratio widens on quieter, less detailed material where prediction has the easiest job.

Can I choose a codec other than MJPEG?

No, and that is not a limitation of the page. A .mjpeg file is defined by its codec: the extension means "these bytes are Motion JPEG", and there is no header in which to declare anything else. That is why MJPEG is the only entry in the Video Codec dropdown here, unlike a container target such as MOV or MKV where the wrapper genuinely is codec-agnostic.

What software opens a raw .mjpeg file?

VLC, mpv, ffmpeg-based utilities and most professional editing and inspection tools read it. Phones, browsers, smart TVs and general-purpose media apps usually do not, because they expect a container with headers and timing rather than a bare frame sequence. If the file has to open somewhere you do not control, this is the wrong target — M2TS to MP4 plays essentially everywhere.

Does the conversion handle the multi-file split that camcorders produce?

Each file is converted on its own, so a recording that the camera split across several M2TS files becomes several .mjpeg files rather than one continuous stream. Because raw MJPEG has no header, though, joining them afterwards is unusually simple — concatenating the files byte-for-byte produces a valid longer stream, since there is nothing at the start or end of each file that needs fixing up. That is not true of MP4 or MOV, where a naive concatenation produces a broken file.

Should I use Quality Preset or Constant Quality?

Constant Quality, on this target. MJPEG is a fixed-quantizer codec and the Quality Preset path is built around scaling a bitrate, a concept the encoder mostly ignores once a quantizer is in force — so leaving the preset alone parks the quantizer at its default and the dropdown does not move the picture the way it does on an H.264 or VP9 target. Switching to Constant Quality exposes the slider that genuinely drives per-frame JPEG quality, and the sizing table above shows what each position costs.

Will 4K M2TS footage work?

It converts, but think hard about whether you want it to. Scaling the per-frame numbers above from 1080p to 3840 × 2160 means roughly four times the bytes per frame, which puts an hour of footage into the hundreds of gigabytes. If you need 4K frames, trim aggressively to the seconds you actually care about, or extract a sparse subset of frames locally with an ffmpeg frame-selection filter instead of converting the whole clip.

What happens to my file after the conversion?

It is uploaded over an encrypted connection, converted on our servers, and both the upload and the .mjpeg output are deleted automatically after a few hours. There is no sign-up, no watermark, and files are never shared or made public. The clip on your own machine is untouched. Because MJPEG output is several times larger than an AVCHD source, the practical limit on a long clip is upload and download time over your connection — trimming first is the single most effective thing you can do.