Multitrack Router — User Guide

Routes one or more selected Sound objects onto virtual mono tracks, places them in time, applies gain and fades, and renders mono, stereo, or multichannel output.

Author: Shai Cohen Affiliation: Department of Music, Bar-Ilan University, Israel Version: 3.3.5 (2026) License: MIT License Requires: Praat ≥ 6.0

Contents:

What this does Quick start Presets Assignment & sequencing Timing & crossfades Overrides syntax Rendering & output Visualization

What this does

Multitrack Router takes all selected Sound objects, maps them to a configurable set of virtual tracks, schedules each track, renders every track as a mono stem, and combines the stems into one final Sound.

5 assignment modes: All on track 1, One track per sound, Round-robin, Manual, and Sequence.
2 timing modes: all segments start at 0, or sequential placement with positive gaps, butt joins, or negative-gap overlaps.
7 presets: Custom plus six routing/timing setups.
Source-channel handling: multichannel inputs are downmixed to mono by default, or channel 1 / channel 2 can be selected with ch:.
Sample-rate unification: segment copies are resampled, when necessary, to the highest sampling frequency among the selected Sounds.
3 outputs: mono sum, stereo equal-power track panning, or one output channel per virtual track.
Schedule and picture: the schedule prints automatically; the picture display is optional.

Signal model: source objects are not modified. Each scheduled audio occurrence is copied, reduced to a mono working segment if necessary, resampled if necessary, gain-scaled, faded, and placed on a mono virtual track. Stereo or multichannel structure is created only when the completed tracks are combined.

Quick start

Select at least one Sound object.
Run Multitrack_Router.praat.
Choose a Preset, or leave Custom to use the visible routing and timing controls directly.
For Manual or Sequence assignment, enter the required syntax in Seq.
Use Overrides only when per-source gain/channel/order or track pan is required.
Choose Mono, Stereo, or Multichannel output; optionally disable normalization or visualization.

Defaults in Custom: 2 tracks, All on track 1, Sequential timing, 0 s gap, 0 s crossfade, 0 dB gain, 0.01 s fade in/out, Mono output, Normalize = yes, Draw_visualization = yes.

Normalization: when enabled, Praat Scale peak: 0.99 is applied whenever the current peak is above 0.001. This is full peak normalization, not only clipping protection, so it can also amplify a quiet result.

Presets

Preset processing occurs after the form and override tags are read. A preset therefore replaces the routing/timing/output fields listed below. Per-source g:, ch:, pan:, and ord: data remain available unless the resulting mode makes a tag irrelevant. Crossfade Montage explicitly forces equal-power crossfade shape.

Preset	Tracks / assignment	Timing	Fades	Output
Custom	Uses form values	Uses form values	Uses form values	Uses form value
Sequential Chain	1 / All on track 1	Sequential, gap 0, crossfade 0	0.005 s in/out	Mono
Crossfade Montage	1 / All on track 1	Sequential, gap −0.05 s, crossfade 0.05 s, equal-power	0 s defaults; crossfade supplies edge fades	Mono
Stereo Spread	2 / Round-robin	Sequential, gap 0	0.01 s in/out	Stereo
Layered Stack	N / One track per sound	All at 0	0.02 s in/out	Mono
Dialogue Assembly	1 / All on track 1	Sequential, +0.30 s gap	0.01 s in/out	Mono
Multichannel Split	N / One track per sound	One segment per track; each starts at 0	0.005 s in/out	N-channel

Assignment & sequencing

Track count

Tracks is clamped to at least 1. In One track per sound, it is forced to the number of selected Sounds. In assignment modes 1–4, a requested track count larger than the number of selected Sounds is reduced to that number. Sequence mode may retain additional empty tracks.

Assignment modes

Mode	Actual behavior
All on track 1	Every selected Sound becomes a segment on track 1.
One track per sound	Sound i is assigned to track i; track count is forced to N.
Round-robin	Sound i goes to `((i−1) mod Tracks)+1`.
Manual	Seq must contain at least N track numbers, one per selected Sound. Each value must be within 1..Tracks.
Sequence	Seq is a pipe-separated list of per-track token sequences. Sound indices may be repeated. `SILx` inserts an explicit silence segment of x seconds.

Sequence example:
1 SIL0.3 2 | 3 4 SIL1.0
Track 1 contains Sound 1, 0.3 s silence, Sound 2. Track 2 contains Sound 3, Sound 4, then 1.0 s silence.

If Sequence mode is selected with an empty Seq field, the script automatically distributes the selected Sound indices round-robin across the requested tracks and uses those generated sequences.

Timing, gaps, overlaps & fades

Timing modes

All at 0: start(segment) = 0 Sequential: next_start = previous_end + Gap_s Gap_s > 0 → silence between adjacent segments Gap_s = 0 → butt join Gap_s < 0 → overlap depth = |Gap_s|

In Sequence mode, explicit SILx tokens are ordinary timed segments, so their durations participate in sequential placement just like audio-segment durations.

Crossfade semantics in v3.3.5

Gap_s and Crossfade_s are independent. A negative Gap_s determines placement and therefore the actual requested overlap. Crossfade_s only sets the edge-fade length inside that overlap and is clamped to the overlap that is actually available.

requested_overlap = |Gap_s| actual_overlap = requested_overlap, except when the start would fall before t = 0 crossfade_length = min(Crossfade_s, actual_overlap) Example: Gap_s = -0.50, Crossfade_s = 0.05 → 0.50 s overlap, with 0.05 s edge fades

If Crossfade_s is shorter than the overlap, the middle of the overlap remains at full level. If it equals the overlap, the fade pair spans the whole overlap. Existing segment fade-in/out values are not shortened: the longer fade wins.

Fade shapes

Ordinary, non-crossfade segment fades use Praat's built-in Hanning Fade in/Fade out.
Segments participating in a crossfade use the explicit shape selected by xf:.
xf:lin uses linear amplitude ramps. For correlated material the amplitude sum meets at unity; uncorrelated material can dip by about 3 dB at the midpoint.
xf:ep uses sine/cosine equal-power ramps. It maintains summed power for uncorrelated material; strongly correlated material can rise by about 3 dB at the midpoint.

Negative fade values are clamped to 0. Fades are applied only to source segments at least 0.03 s long, and each fade is capped at 45% of that segment's duration.

Overrides syntax

Overrides is a compact tag field. Tags are separated by whitespace; each tag's values continue until the next tag.

Tag	Meaning	Rules
`g:`	Per-source gain in dB	Values correspond to selected Sound indices. Missing entries retain Gain_dB; extra entries beyond N are ignored.
`ch:`	Per-source channel choice	`0` = convert multichannel source to mono, `1` = source channel 1, `2` = source channel 2. Invalid values are reset to 0. Mono sources remain mono.
`pan:`	Per-track stereo pan	Only used for Stereo output. Values are clamped to −1..+1. If a manual list is present but shorter than the track count, unspecified tracks remain centered rather than being auto-spread.
`ord:`	Per-source order rank	Requires at least N values. Lower rank is placed earlier within the source's assigned track. It is not a list of sound indices and is not applied to explicit Sequence mode.
`xf:`	Crossfade shape	`lin` selects linear; every other value selects equal-power. Crossfade Montage forces equal-power.
`sched`	Backward-compatible no-op	The schedule already prints by default in v3.3+.

Examples

g:-3 0 -6 → Sound 1 = -3 dB, Sound 2 = 0 dB, Sound 3 = -6 dB ch:0 2 1 → Sound 1 downmix-to-mono, Sound 2 channel 2, Sound 3 channel 1 ord:3 1 2 → ranks are [3,1,2], so on the same track the order becomes Sound 2 → Sound 3 → Sound 1 pan:-1 0 1 → tracks 1/2/3 hard-left / centre / hard-right

Rendering & output

Per-segment rendering

Copy the source Sound.
If it has more than one channel, apply the ch: choice: channel 1, channel 2, or Convert to mono.
If its sampling frequency differs from the highest selected-source rate, resample the copy to that rate with precision 50.
Apply gain 10^(dB/20).
Apply fade-in/out.
Add only the affected sample range into the scheduled position on the mono track canvas.

Output formats

Output	Combination	Object name
Mono	All mono track stems summed directly.	`Routed_mono`
Stereo	Each track is sent to L/R by equal-power panning.	`Routed_stereo`
Multichannel	Track t becomes output channel t.	`Routed_multi`

Stereo panning

angle = (pan + 1) / 2 × π/2 L = cos(angle) R = sin(angle) pan = -1 → hard left pan = 0 → centre (≈0.707 / 0.707) pan = +1 → hard right

Without a pan: override, only active tracks are spread evenly from −1 to +1. If only one track is active, it is centered.

Duration and normalization

Each virtual track canvas extends to its latest segment end plus a 0.0005 s tail margin. Final duration is the maximum track duration. No limiter is used.

If Normalize = yes and the pre-normalization peak is above 0.001, the complete output is peak-normalized to 0.99 with Praat Scale peak. With Normalize disabled, summed peaks may exceed ±1 internally.

Schedule

When at least one segment exists, the Info window automatically prints one line per segment with segment number, track, AUDIO/SILENCE type, source index, start/end times, gain, stored fade-in/out values, and source name. Render-time fade caps are not written back into those stored schedule values.

Playback: the completed output is always selected and played at the end of the script. There is no Play-result toggle in v3.3.5.

Visualization

With Draw_visualization = yes, the script draws an 8-wide Praat picture with three main regions:

Track timeline: up to the first 8 tracks. Segment width represents time; audio-block height is scaled from the source RMS; source index controls block colour. Explicit silence is shown as a low neutral block.
Timing overlays: fade-in/out lines are drawn on audio blocks; overlap pairs that have a crossfade are hatched. For stereo output, each visible track also has an L/C/R pan indicator.
Output waveform: mono shows one waveform; stereo overlays L in blue and R in orange; multichannel shows channel 1 as a representative waveform.
Summary: preset, assignment, layer/sequential mode, gap, crossfade value/shape, default fades, output format, duration, peak, gain, segment count, and normalization state.

The timeline is a process view, not a sample-accurate editing display. Its fade lines are simple visual approximations of the envelopes, and only the first eight tracks are shown when more tracks exist.

Practical uses

Sequential assembly

Place phrases, events, or dialogue one after another with explicit gaps or overlaps.

Layered construction

Start sources together, keep them on independent virtual tracks, then sum or spatialize the tracks at output.

Multichannel preparation

Use one track per source and Multichannel output to produce one output channel per virtual track.

Algorithmic sequencing

Use repeated source indices and SILx markers in Sequence mode to construct track-specific event lists.

Notes & edge cases

Manual assignment: the Seq list must provide at least one valid track number for every selected Sound.
Sequence syntax: sound tokens must resolve to indices 1..N; SILx requires a strictly positive numeric duration.
Ordering: ord: is a rank vector, not a requested sequence. Equal ranks retain their existing relative order.
Crossfade without overlap: Crossfade_s does not create an overlap by itself. A negative Gap_s is required.
Multiple-channel sources: every source occurrence becomes mono before track rendering; use ch: when a specific first or second source channel is required.
Normalization: disable it when preserving absolute level relationships at the final output is important.