Initializing... drag & drop files here
Supports: AVCHD
An .avchd file is high-definition camcorder video — H.264 wrapped in an MPEG transport stream with a Dolby AC-3 soundtrack, the format Sony and Panasonic introduced in 2006. A .3g2 is the 3GPP2 multimedia file format from 3GPP2 C.S0050, the recording and playback container CDMA2000 handsets used, sized for the bandwidth and screens of 2005.
Converting one to the other is a deliberate downgrade, and there are only a few reasons to want it: an ingest pipeline that accepts .3g2 and nothing else, a device or emulator from that generation, or a hard size budget where 3GPP2's low-bitrate profiles are exactly what you need. This page is written so you make that downgrade on purpose rather than by accident — because two of the settings here can quietly do far more damage than the conversion itself requires.
Video Codec lists H.263, H.264, MPEG-4 and Xvid, and opens on H.264. Leave it there unless something specific demands otherwise, because H.263 behaves in a way nothing else on the page does.
| Video Codec | What happens to a 1920 x 1080 AVCHD frame | Aspect ratio |
|---|---|---|
| H.264 (preselected) | Stays 1920 x 1080 unless you change Video resolution | Preserved |
| MPEG-4 / Xvid | Follows your Video resolution setting, rounded for the codec | Preserved |
| H.263 | Forced to 704 x 576, whatever Video resolution says | Squashed — 16:9 is stretched into 11:9 |
H.263 accepts only five picture sizes — 128 x 96, 176 x 144, 352 x 288, 704 x 576 and 1408 x 1152 — and the encoder hard-fails on anything else. Ask FFmpeg directly to encode 1920 x 1080 as H.263 and it answers "The specified picture size of 1920x1080 is not valid for the H.263 codec" and stops. Rather than fail the job, the pipeline snaps the frame to the largest legal size that fits inside the source: for any HD footage, 1080-line or 720-line, that is 704 x 576. The snap sets both dimensions directly, so a widescreen frame is not letterboxed into that box — it is squeezed into it, and everyone in shot gets noticeably narrower.
That is a genuine reason to pick H.263 anyway if you are feeding a period-correct handset that decodes nothing else. It is a bad surprise if you picked it thinking it was "the phone codec". H.264 in a .3g2 is entirely legal and is what later 3GPP2 devices expected.
.avchd onto the page or click "+ Add Files" to browse. Camcorder files are large, so the transfer is usually the slow part; several can be queued with one set of settings..3g2. Files are uploaded over an encrypted connection, re-encoded on our servers and deleted automatically after a few hours — no sign-up, no watermark, never shared or made public.AVCHD carries Dolby AC-3, up to 5.1 channels. The 3GPP2 container does not allow AC-3 at all, so the audio is always decoded and re-encoded into one of the three codecs this page offers — and two of them are speech codecs with fixed operating points:
| Audio Codec | Sample rate | Channels | Suited to |
|---|---|---|---|
| AAC | Follows your Audio setting or the source | Stereo or mono | Anything you want to listen to; the sensible default |
| AMR Narrow Band | Forced to 8,000 Hz | Forced to mono | Speech only, and maximum compatibility with old CDMA handsets |
| AMR Wide Band | Forced to 16,000 Hz | Forced to mono | Speech with a little more clarity, on devices that support it |
Those AMR constraints are absolute — pick AMR Narrow Band and the encode lands at 8 kHz in a single channel no matter what the camcorder recorded or what you set elsewhere. A 5.1 concert recording becomes a mono telephone call. Choose AAC unless a device genuinely requires AMR.
You selected H.263 as the Video Codec. H.263 supports only five fixed picture sizes, and the pipeline snaps your frame to the largest one that fits inside the source — for any HD camcorder footage that is 704 x 576, an 11:9 box. Because both dimensions are set directly, a 16:9 frame is compressed horizontally rather than letterboxed, which makes everything look tall and narrow. Switch the codec to H.264, which imposes no such restriction, and the picture keeps its proportions.
.3g2 actually valid?Yes. The 3GPP2 file format is built on the same ISO base media structure as MP4, and later revisions of the specification added H.264 alongside the earlier H.263 and MPEG-4 Part 2 video. Devices from the mid-2000s onward generally handle it. The one situation where H.264 is wrong is a genuinely old handset or emulator that only ever implemented H.263 — and that is the case where the frame-size squash is a price worth paying.
It could not travel. The 3GPP2 container does not permit AC-3, so the AVCHD soundtrack is decoded and re-encoded into AAC, AMR Narrow Band or AMR Wide Band. AAC can carry stereo and will preserve a reasonable amount of a 5.1 mix once it is downmixed; both AMR options are mono-only speech codecs and will discard everything a surround mix was for. If keeping the Dolby track intact matters more than the container, extract it separately with AVCHD to AC3 and keep it alongside.
Usually yes, if the file is going anywhere the container implies. 3GPP2 exists because handsets had small screens and slow connections; a 1920 x 1080 .3g2 is legal but defeats the purpose, and no device that wants this format will have a display to match. Set Video resolution to a Preset Resolution — the presets set the height and let the width follow the source aspect, so 480p on 16:9 footage gives you roughly 854 x 480. If you have chosen H.263, the resolution setting is overridden by the frame-size snap regardless.
That depends far more on the bitrate you choose than on the container. AVCHD camcorders record at roughly 17 to 24 Mbps for 1080-line footage, so almost any setting here is a large reduction. The most predictable route is Specific file size: enter a target and let the pipeline work out the bitrate from the clip's duration. Constant Bitrate is the next most predictable. The Quality Preset is fine when you have no particular target, and on H.264 it maps to a quality level rather than to a rate.
No. Much AVCHD material was recorded interlaced — 1080i in particular — and there is no deinterlacing step anywhere in this pipeline. Combing on fast motion is decoded as it was recorded and encoded straight back into the output, and shrinking the frame makes it look worse rather than better because the interleaved fields get blended into one another. If your footage is interlaced and the result matters, deinterlace it in a desktop editor before uploading.
.3g2 and .3gp?They are sibling containers on the same ISO base media foundation, split by network family. 3GPP standardised .3gp in TS 26.244 for GSM and UMTS phones; 3GPP2 standardised .3g2 in C.S0050 for CDMA2000 networks, and it serves under the MIME type video/3gpp2. The practical difference is which audio codecs each was permitted to carry — .3g2 also allowed the CDMA vocoders. If your target actually wants the GSM-side file, use AVCHD to 3GP instead.
.avchd will not upload or decode. What is wrong?Check what you actually have. Camcorders write AVCHD as a directory structure — a BDMV or AVCHD folder containing a STREAM directory of .mts or .m2ts files — not as a single .avchd file, so a file with that extension is usually something a tool renamed. If your footage is still on the card or in a folder, upload the individual stream file instead and use MTS to 3G2. A partially copied file decodes up to the damaged point and stops, which shows up as an output shorter than the source.
Yes, and on camcorder footage it is worth doing. Set Trim to Time Range with a Start time and a Duration and only that section is decoded and encoded. If you would rather cut before uploading — which also saves you transferring the whole recording — Video Cutter accepts .avchd directly, and Trim AVCHD writes the trimmed result back in the same format.
It is uploaded over an encrypted connection, re-encoded on our servers, and both the upload and the .3g2 it produced are deleted automatically after a few hours. Nothing is shared or made public, and there is no sign-up or watermark. Because AVCHD files are large and the output is small, transfer time dominates — trimming first, or converting in modest batches, is the fastest way through a shoot's worth of footage.