Audio Transcoding
Audio transcoding decodes a source presentation and encodes a new audio stream for delivery or storage.
A movie's video can fit a client while its soundtrack does not. A server can keep the video and construct a new audio presentation. Audio transcoding is the decode/process/encode part of that route.
The operation is easier to assess when its intermediate representation is made visible.
A 7.1 source converted to stereo AAC
Consider a hypothetical 48 kHz 7.1 source and a client requiring stereo AAC:
| Stage | Representation |
|---|---|
| Source reading | Coded eight-channel presentation and timestamps |
| Audio decoding | Reconstructed samples with eight named roles |
| Layout conversion | Two-channel result incorporating the appropriate source contributions |
| AAC encoding | New coded stereo approximation |
| Delivery packaging | Compatible video and new timed audio in the target package |
The audio decoder and encoder are different jobs. The decoder reconstructs what the source represents. The encoder analyzes its supplied PCM and chooses a new representation that meets its settings.
Emby's Direct Stream description includes compatible video retained while audio or packaging changes. The overall delivery label does not imply the source soundtrack's layout or codec remained unchanged.
Separate conversions inside the pipeline
Downmixing changes spatial channels. Resampling calculates another time grid. Gain or dynamics processing changes amplitudes. The final encoder changes the coded representation.
These can occur together, but each has its own preservation effect:
| Operation | Consequence |
|---|---|
| Exact supported lossless decode/re-encode, no intervening changes | Can preserve the samples |
| Downmix | Combines independent channels |
| Resample | Changes sample positions/count while retaining intended time |
| Lossy encode | Adds its representation's approximation |
| Ordinary channel encode of an immersive result | Does not retain original objects merely by using the same codec family |
An encoder generally also introduces buffering, delay, and final padding. The delivery's timing must account for those so the new soundtrack remains aligned and has the intended audible boundaries.
Distinguish a new encoding from remuxing
A remux reads coded packets and writes another supported package, translating necessary configuration and timing. It can leave the coded audio intact.
A transcode instead reconstructs the signal and invokes an encoder. Moving AAC from ADTS into a compatible MP4 arrangement may be a framing change; changing AAC into another codec is a reconstruction-and-encoding operation.
Understand generation loss at the right boundary
If the source is already lossy, the new encoder receives that reconstruction. A larger target budget can avoid adding much more error, but cannot recover the source details previously merged or discarded.
A lossless source avoids that prior approximation, yet a subsequent lossy delivery still makes a new approximation. Keeping the source master allows future sessions to start from it rather than repeatedly converting intermediate copies.
Verify what the destination received
Compare source and delivered codec, layout, rate, and timeline. Then check whether the output added further processing. For missing dialogue, inspect the layout conversion; for a timing shift, inspect delay compensation; for altered texture, inspect coding generations and settings.
This makes a transcoding report about identifiable operations and outcomes, rather than treating “conversion happened” as a complete explanation.