Transcoding is the process of decoding encoded video and re-encoding it with a different codec, bitrate, resolution, frame rate or other encoding parameters, usually producing a new file or stream. In video workflows built around containers and formats, it is distinct from remuxing, which moves the existing encoded streams into a different container without touching the compressed data.
Transcoding is unavoidable in modern delivery: a single camera master is typically transcoded into editing proxies, a mezzanine master and a ladder of streaming renditions. Because most delivery codecs are lossy, every transcode is a generation of quality loss that must be planned rather than repeated casually.
How transcoding works
Transcoding works as a pipeline of stages: the container is demuxed into its elementary streams, the video stream is decoded to uncompressed frames, those frames are optionally processed, and the result is encoded again and muxed into a target container. Audio, subtitles and metadata follow their own paths through the same pipeline.
| Stage | What happens | Typical concern |
|---|---|---|
| Demuxing | Container split into video, audio and data streams | Correct track selection |
| Decoding | Compressed frames expanded to raw pixels | Codec support, decode speed |
| Processing | Scaling, deinterlacing, frame rate conversion, color conversion | Filter quality, color accuracy |
| Encoding | Raw frames compressed with target codec | Bitrate, GOP, quality settings |
| Muxing | Streams written into target container | Timestamps, compatibility |
The processing stage is where many quality problems arise. Downscaling filters, chroma subsampling changes and color space conversions, for example between Rec. 709 and Rec. 2020, must be configured explicitly, because defaults are not always correct for the source.
Audio is often handled differently from the picture. When the source audio already uses a codec the target container accepts, such as AAC in MP4, it can be copied without re-encoding while only the video is transcoded. Uncompressed PCM audio from a camera, by contrast, is usually encoded to AAC or Opus for delivery to save space.
Transcoding, remuxing and related operations
Several operations change a video file, but only some of them re-encode the picture. The distinction matters because re-encoding costs time and quality, while container-level changes are fast and lossless.
| Operation | Re-encodes video | Quality effect | Typical example |
|---|---|---|---|
| Transcoding | Yes | Generation loss with lossy codecs | ProRes master to H.264 for web |
| Transrating | Yes, same codec | Loss from lower bitrate | High-bitrate H.264 to lower bitrate H.264 |
| Transsizing | Yes, new resolution | Loss of detail when downscaling | 4K to 1080p |
| Remuxing | No | Lossless | MKV to MP4 with the same H.264 stream |
| Rewrapping | No | Lossless | MXF camera file to MOV for an editor |
Remuxing is only possible when the target container supports the existing codec. An H.264 stream moves easily between MKV, MP4 and MPEG-TS, whereas a codec that the target container or player does not support forces a full transcode.
Common transcoding scenarios
Most professional workflows contain several planned transcodes, each with a different goal. Choosing the right target codec for each step prevents unnecessary loss and keeps files manageable.
| Scenario | Source | Target | Goal |
|---|---|---|---|
| Editing proxies | Long-GOP H.265 camera files | Low-resolution ProRes Proxy or DNxHR LB | Smooth timeline playback |
| Mezzanine master | Edited timeline | ProRes 422 HQ or DNxHR HQX | High-quality source for all deliverables |
| Streaming ladder | Mezzanine master | H.264, H.265 or AV1 at several bitrates | Adaptive delivery to all devices |
| Broadcast delivery | Mezzanine master | Codec and wrapper required by the broadcaster | Technical compliance |
| Archive conversion | Legacy formats such as MPEG-2 | Modern mezzanine or lossless codec | Long-term accessibility |
A streaming ladder is a set of renditions of the same content at increasing resolution and bitrate, for example from 360p for weak mobile connections to 2160p for fast broadband. Each rung is a separate transcode from the same mezzanine source, never from another rung, to avoid compounding generation loss.
Key settings and hardware choices
Transcoding quality and speed depend on rate control, encoder preset and hardware. Software encoders usually achieve the best quality per bit, while hardware encoders such as NVENC and other GPU or system-on-chip encoders offer much higher speed at a modest efficiency cost.
| Setting | Common options | What it affects |
|---|---|---|
| Rate control | Constant quality, VBR, CBR | Bitrate predictability versus consistent quality |
| Encoder preset | Fast to very slow | Encoding speed versus compression efficiency |
| GOP length | 1 to 10 seconds | Seeking, streaming segment alignment |
| Bit depth | 8-bit or 10-bit | Banding in gradients, HDR support |
| Hardware encoding | Software or GPU encoder | Speed, power use, efficiency |
Matching settings to the destination is essential. A constant bitrate suits live broadcast links with fixed capacity, while constant quality or variable bitrate suits files and on-demand streaming where size can fluctuate.
Transcoding is also computationally heavy. Encoding a long 4K program with a slow software preset can take many times the program’s running time, so large platforms distribute work across many machines by splitting files into chunks, encoding them in parallel and joining the results at GOP boundaries.
Generation loss and common mistakes
Every lossy transcode discards information permanently, and errors accumulate across generations. Artifacts such as blocking, banding and smeared texture grow with each pass, which is why experienced engineers transcode from the highest-quality source available.
- Transcoding when remuxing would do: changing only the container should never trigger a re-encode.
- Chaining delivery files: encoding a web H.264 file again for another platform stacks losses; return to the master.
- Upscaling bitrate on a poor source: a higher bitrate cannot restore detail lost in an earlier transcode.
- Ignoring color metadata: wrong color primaries or range flags cause washed-out or crushed images.
- Mixing frame rates carelessly: converting 25p to 30p without proper handling creates judder.
Stripped to the essentials
Video files are packed in different compression styles, much like documents saved in different file types. Transcoding means unpacking the video completely and packing it again in a new style. You might do it so the video plays on a phone or uploads faster.
It is a bit like photocopying a photocopy. Each round loses a little quality, so professionals always make new copies from the best original they have.
Where it is applied
A small production company finishes a 20-minute documentary edited from 4K H.265 camera files. During editing, the editor worked from ProRes Proxy transcodes at 1080p, because the long-GOP camera files stuttered on the timeline. After picture lock, the editor relinks to the original camera files and exports a ProRes 422 HQ mezzanine master.
From that single master, the company transcodes a festival screening file, a broadcaster deliverable in the required MXF format and a streaming ladder of H.264 renditions from 480p to 2160p. When a platform later rejects a subtitle track in an MKV file, the team simply remuxes to MP4, leaving the picture untouched and avoiding another round of quality loss.