BitrateCraft · Audio and text tracks · bitratecraft.org

Disclaimer: Independent technical documentation and configuration guide. Not affiliated with The HandBrake Team. No software downloads hosted

Audio Encoders, Mixdown and Routing

In HandBrake the audio configuration is a per-track decision list rather than one global setting. Each row in the audio panel binds a source track to an encoder family, a rate or quality target, a channel layout and a gain offset, and the rows are written to the output in the order they appear. Treating those four properties as independent switches is what makes an audio layout predictable across players, receivers and portable devices.

How a track row is resolved

A row is evaluated at job start in three stages. The source track is first decoded to a linear PCM buffer at its native sample rate. The mixdown matrix then folds that buffer into the requested channel layout, applying the relevant coefficients and any normalisation the operator has enabled. Finally the selected encoder consumes the result while the muxer interleaves the finished frames with the video stream.

Failures are quiet by design. If the matrix expects six channels and the source holds two, the row falls back to the closest layout it can satisfy, so an ambitious surround request can end up written as plain stereo with no error shown. Track order matters for the same reason: the first row is flagged as the default in the container header, and any player that ignores language tags will select it. Put the broadly compatible mix first and the specialised program material second.

Encoder families compared

Each family trades efficiency, compatibility and processing cost differently. The table summarises the practical envelope rather than the theoretical maximum.

Container compatibility usually decides the shortlist. A lossless row inside MP4 may be accepted by the muxer and still be skipped by a television that expects only AAC. When an output must play everywhere, a lossy compatibility row in first position covers the requirement without discarding the richer tracks behind it.

Bitrate mode and quality mode

Every lossy row offers a rate target, and several families also expose a quality control. A rate target asks the encoder to hold an average across the whole program, which is predictable for planning but blunt: quiet dialogue and dense music are billed at the same rate. Quality mode lets the encoder spend bits where the signal demands them, which is normally the better choice for archival work whose loudness and spectral content vary between chapters. Fixed targets remain useful when the result has to fit a known channel or a fixed storage envelope.

Family behaviour differs. AAC responds smoothly across its quality range and degrades gracefully below 96 kbps per stereo pair. Opus is efficient well below that point and is the sensible choice for speech-heavy material. AC-3 should be left inside its established rate bands, because decoders in older amplifiers accept the presence of the track only when those bands are respected.

Routing map from source to muxer

The diagram below shows how a two-row configuration reaches the output file. Each source track is decoded once, folded by its own matrix, and encoded independently.

Channel mixdown

Mixdown decides how many channels the encoded row carries and which coefficients map the source channels into them. The common targets are mono, stereo, Dolby Pro Logic II, 5.1 and 7.1, plus a pass-through option that copies multichannel content without folding it. Pass-through preserves the original bit-for-bit where the family allows it and costs almost nothing in processing, but it also inherits the source quality, including any artefacts already present.

Summing multiple channels into fewer ones multiplies the risk of clipping, because six correlated sources can add well above the headroom of a single channel. The normalisation setting exists for that case: it scales the folded result down so the loudest sample stays inside range. It costs a little level, and for material that has already been mastered close to full scale it is the difference between a clean result and audible flat-topping.

Sample rate handling

Sample rate belongs to the output row, not to the source, so a row can resample even when the material is being copied without a mixdown change. Video-derived sources are almost always 48 kHz, and keeping that rate avoids a conversion step and its filter cost. The 44.1 kHz target belongs mainly to music-derived material, while 32 kHz is a legacy television rate and 96 kHz suits archival captures where the extra bandwidth is genuinely present. Resampling does not add information; it only spends processor time and can introduce ringing at the band edge if the ratio is awkward.

Gain and dynamic range

The gain control offsets the row in decibels before encoding. Positive values of roughly three to six decibels recover material mixed for cinema playback; larger values rapidly push peaks into the limiter. Dynamic range compression, where the family supports it, works on the metadata that some decoders honour, so its audible effect depends on the playback chain rather than on the encoded samples themselves. Apply gain first and check the meter before trusting a large offset.

Routing rules worth keeping

1. Decide the compatibility row first and place it at the top of the panel.
2. Match every mixdown to the smallest source channel count the title actually offers, not the largest in the batch.
3. Add surround or lossless rows only when the material justifies them.
4. Keep sample rate uniform across all rows so that downstream players never switch streams mid-playback.
5. Apply gain before the matrix when a fold is involved, so the limiter sees the summed signal.
6. Audit the fallback chain once per batch by checking one short sample of each title class.

Questions that come up

Related reading: the quality and rate factor guide explains how the video half of the same job is configured, and the transcoding glossary defines the mixdown and matrix terms used above.