The Vital preset file

This page states what is known about the .vital preset file. No official format document exists. The facts come from the open-source snapshot of Vital, from renders of its open engine and from sampled presets.

Every row names a build, a status and a source.

Builds. source 1.0.6 is the open-source snapshot at commit 636ca0e. open engine is that engine compiled into the vita 0.1.0 Python binding, which reports itself as 99999.9.9. A retail build is named by its version, such as 1.0.7.
Statuses.
- checked: read in the source.
- verified: confirmed by a probe or a load check of ours on the named build.
- reported: an outside source says so.
- open: unknown.
- untestable by rule: a check that needs a step the project never takes, such as running the app.

A row verified on the open engine is not verified on retail. The pinned retail build is stable 1.0.7. The early-access build is 1.6.4 (reported). The checks still open are listed under Open items.

File and top-level keys

Fact	Build	Status	Source
A preset is one compact JSON text.	source 1.0.6	checked	`load_save.cpp`
The top level holds `synth_version`, `preset_name`, `author`, `comments`, `preset_style`, `macro1` to `macro4` and `settings`.	open engine, 1.0.7	verified	probe: saved default states
`settings` holds 772 numeric controls and four structured parts: `sample`, `modulations`, `wavetables` and `lfos`.	open engine, 1.0.7	verified	probe: saved default states
The app saves the keys of each object in alphabetical order.	1.0.7	verified	load check: the re-save of `Bass`
A re-save keeps every control within one 32-bit float step and rewrites the `version` of each wavetable to the app’s build. A load and save rewrites the sample: our noise sample came back 4 frames later with some values one step off, so compare a re-saved sample by its name, length and rate, not its bytes.	1.0.7	verified	load check: the re-save of `Bass`; the open engine keeps the sample
A file without `synth_version`, `settings` or `settings.sample`, or a sample without `length`, `sample_rate` or `samples`, makes the app throw.	source 1.0.6	checked	`load_save.cpp`
A missing control takes its default.	open engine	verified	probe: `minimal.vital` reads back 772 controls, each at the fresh engine’s default
An unknown key loads and is dropped on save, in `settings` and at the top level.	open engine	verified	load check: the unknown-keys variant
A file without `preset_name` and `author` loads.	open engine	verified	load check: the no-name variant
An out-of-range control is applied as it is. The loader does not clamp.	source 1.0.6	checked	`load_save.cpp`
Factory presets hold values outside the table range, such as a dry/wet above its maximum or the 32-bit float form of a bound. `preset apply` warns about such a value in a key no edit touched and refuses it in an edited key.	1.0.x	verified	observed in factory presets of Vital 1.0.x
A text in a control fails the load. A bool loads. The writer of this project writes neither.	open engine	verified	probe: `filter_1_cutoff` set to `"x"` and to `true`
The app overwrites `preset_name` with the file name on open.	source 1.0.6	checked	`synth_base.cpp`
The writer of this project prints each number as the shortest text that reads back to the same value. `1E5` becomes `100000.0`. Values stay bit for bit.	this project	verified	unit test `number_text_is_canonical_and_values_are_kept`

Version gate

The app compares the major.minor part of synth_version with its own build. The patch number is ignored.

Case	Build	Status	Source
A file whose major and minor are newer is rejected with a dialog.	source 1.0.6	checked	`load_save.cpp`
A file one patch newer loads.	source 1.0.6	checked	`load_save.cpp`
An older file runs about 30 upgrade steps before it loads. Files from 1.0.0 on need none.	source 1.0.6	checked	`load_save.cpp`
The plugin state path ignores a rejected file. The patch keeps its last sound, and no dialog appears.	source 1.0.6	checked	`synth_plugin.cpp`
The stamps `1.0.6`, `1.0.9`, `1.1.0`, `1.5.5` and `0.9.0` load. `100000.0.0` and a text that is not a version are rejected.	open engine	verified	probe 5 and the stamp fixtures
A file stamped `1.0.6` loads, with every edited control in place.	1.0.7	verified	load check: `Bass`
Files stamped `1.5.5`, `1.6.0` and `1.7.0` are refused with the dialog “Preset was created with a newer version”.	1.0.7	verified	load check: the stamp variants
The app writes its own build, `1.0.7`, as the stamp on save.	1.0.7	verified	load check: the re-save of `Bass`
Presets saved by 1.6.x do not open in 1.5.5.	1.5.5	reported	Vital forum
Which stamps the early-access 1.6.4 accepts.	1.6.4	open	optional forward check

Stamp rule. Files built from the built-in template carry 1.0.6, the open-source build, as DEFAULT_STAMP. Stable 1.0.7 loads them: the gate compares only major and minor, and files from 1.0.0 on run no upgrade step. preset apply keeps the stamp of its template and never lowers it. A template older than 1.0.0, or one that holds sub_octave, is refused. For the early-access builds, pass a template saved by your own app with --template.

Retail-only keys

Retail builds write keys that the open source does not know.

Key	Type	First seen	Default	Status	Source
`custom_warps`	list of line-generator states, one per oscillator	1.5.3	open	checked	8 of 24 sampled presets
`random_values`	list of `{seed}` objects, one per random LFO	1.5.3	open	checked	the same sample
`osc_1_spectral_morph_phase` to `osc_3_`	number	1.5.3	open	checked	the same sample

Stable 1.0.7 writes none of them. It loads a file that holds them, and so does the open engine, which drops them on save (verified, load check: the retail-keys variant). preset inspect marks them. Whether builds 1.6.x add more keys is open.

Template

Fact	Build	Status	Source
The built-in template holds all 772 controls at their registry defaults, 64 empty modulation slots, three sawtooth wavetables, eight triangle LFO shapes and a seeded noise sample.	this project	checked	`src/preset/vital/template.rs`
Its key set and every default match a fresh open engine and the saved default state of 1.0.7.	open engine, 1.0.7	verified	probe 8: 0 keys differ
It loads, and every control reads back unchanged.	open engine	verified	load check: the template variant
The built-in template is not the app’s Init preset. Its wavetables and sample are our own.	this project	checked	`src/preset/vital/template.rs`

Parameters and laws

The registry in src/preset/vital/registry.json holds 794 controls. 22 are legacy keys that no module creates. Our own tool extracts the names, ranges, defaults, engine scales and labels from the source; the units, laws and provenance are ours.

Fact	Build	Status	Source
The open engine and 1.0.7 create exactly the 772 non-legacy controls.	open engine, 1.0.7	verified	probe: key sets
The 1.0.7 default state holds all 772 controls at the registry default. These carry `verified_on = 1.0.7`, which confirms the key and the default only.	1.0.7	verified	load check: the 1.0.7 default state
The open engine reports the same range, default and first build as the registry for 763 controls; the other 9 are listed below. The registry evidence says which.	open engine	verified	probe: control details
A filter cutoff follows $f = 8.1758 \cdot 2^{s /12}$ Hz: stored 66.86 shows about 390 Hz.	1.0.7	verified	probe: a UI edit read back from the saved file
Display laws read and typed in the value box, each matching the registry law: `portamento_time` is $2^{s}$ seconds (0.5 s stores -1), `osc_1_unison_detune` is $s^{2}$ percent (25 % stores 5), `filter_1_cutoff` shows $s - 60$ semitones (12 stores 72), `distortion_drive` is $s$ dB (12 stores 12). The registry tags these four `measured`, with `verified_on = 1.0.7`.	1.0.7	verified	probe: values typed in the app and read back from the saved file
A route amount shows in the destination’s units, as the amount times the destination’s stored span. On `filter_1_cutoff`, span 128, 64 semitones stores 0.5. `preset apply` converts a semitone amount the same way.	1.0.7	verified	probe: a value typed in the app; unit tests of the edit list
An LFO tempo menu stores the index of its label: `2/1` is 5 and `1/4` is 8.	1.0.7	verified	probe: a UI edit read back from the saved file
The macOS bundles of 1.0.7 are x86_64 only, checked with `lipo -archs` on 2026-10-26. A plugin host running natively on arm64 cannot load them. On Apple silicon an x86_64 app runs under Rosetta.	1.0.7	verified	probe 1; Rosetta is a platform fact
`delay_tempo` and `delay_aux_tempo` show the menu label at the stored value itself. Value 4 is `4/1`. The `vita` binding counts from the range start instead and shows `Freeze`.	source 1.0.6	checked	`synth_slider.cpp`; the binding differs
`filter_*_style` has values 0 to 9 and five labels.	source 1.0.6 and open engine	verified	probe: control details
`osc_*_destination` and `sample_destination` allow 0 to 14, but only 0 to 13 have labels. The open engine reports 13 as the top.	source 1.0.6	checked	`synth_parameters.cpp`; the binding differs
An oscillator level is the square of its stored value: $g = s^{2}$ .	open engine	verified	render: halving the stored level quarters the amplitude
The outer unison offset is `detune_range` times `unison_detune` squared, in cents.	open engine	verified	render: range 2 and detune 10 give 200 cents
Unison voice offsets follow `laws::unison_offsets` for 4 and 8 voices at powers 0 and 1.5, within 0.03 cents. Odd counts are not rendered.	open engine	verified	render, unit test `offsets_match_the_render`
Unison gains follow `laws::unison_gains` for 3, 4, 5 and 8 voices, within 0.001 of the single-voice amplitude. A gain belongs to a pair of voices, which share it equally.	open engine	verified	render, unit test `gains_match_the_render`
The voice gain is $(e \cdot lerp (1, v, t) \cdot a)^{2}$ , with envelope $e$ , velocity $v$ , velocity track $t$ and `voice_amplitude` $a$ .	open engine	verified	render, unit test `the_amp_gain_squares_like_the_render`
Laws tagged `interim` in the registry: resonance, filter drive, filter shape, distortion except `distortion_drive`, compressor, EQ, effect feedback, depth and mix, reverb, oscillator warp, curve powers, and the controls with no display unit.	source 1.0.6	checked	`src/preset/vital/registry.rs`; each carries a tuning note

The interim laws wait for the render calibration. They are not settled here.

Modulation, LFO shapes, wavetables and names

Fact	Build	Status	Source
`modulations` holds 64 `{source, destination}` entries. A slot’s amount, bipolar flag, power, stereo and bypass are the controls `modulation_N_*`.	open engine	verified	probe: a fresh engine’s saved state
The 32 source names. A route from each of them loads.	open engine, 1.0.7	verified	load check: the every-source variant; in 1.0.7 the matrix shows 32 rows, and the value box reads 12.8 for amount 0.1 into the cutoff
A route amount is a fraction of the destination’s stored range: $Δ = (x - 0.5 b) \cdot amount \cdot (max - min)$ .	source 1.0.6	checked	`laws::modulation_delta`
Routes target numbers, counts, menus such as `osc_1_transpose` and `lfo_1_tempo`, and the amount or power of another slot, never a switch. `preset apply` refuses a switch and the bipolar, stereo and bypass flags of a slot.	1.0.x	verified	observed in factory presets of Vital 1.0.x
A Wave Source holds `name`, `author`, `version`, `remove_all_dc`, `full_normalize` and `groups`. A component holds `type`, `interpolation`, `interpolation_style` and `keyframes`. A keyframe holds `position` and `wave_data`.	open engine	verified	probe 9: the engine’s saved state has the same fields as ours
`wave_data` is base64 of 2048 little-endian 32-bit floats: 10924 characters. Positions run from 0 to 256.	source 1.0.6	checked	`wave_source.cpp`
Tables of 1 and of 257 keyframes load. One keyframe shows a single frame; with 257 the frame position sweeps the wave.	open engine, 1.0.7	verified	load check: the wavetable variants
An LFO shape holds `name`, `num_points`, `points`, `powers` and `smooth`, with one power per point. A shape of 100 points loads.	open engine, 1.0.7	verified	probe 9 and the 100-point variant
A stored point y is drawn upside down: the output is $1 - y$ , so a stored 0 is the top. The default triangle stores $(0, 1), (0.5, 0), (1, 1)$ and peaks at the centre. `LfoShape` takes outputs and stores $1 - y$ .	1.0.7	verified	load check: a shape of rising stored values showed falling saws
`effect_chain_order` codes the effect list, top to bottom. Code 0 is Chorus, Compressor, Delay, Distortion, EQ, Filter, Flanger, Phaser, Reverb. Number the effects in that order. Each effect after the first adds a mixed-radix digit, the count of effects above it with a higher number, in base its position plus one. The first digit is the most significant: for $i$ from 1 to 8, $code = code \cdot (i + 1) + digit_{i}$ . The codes run from 0 to $9! - 1$ .	1.0.7	verified	probe 10: lists rearranged in the app saved 260650, 8 and 181440; unit tests `the_effect_order_codes_match_vital` and `every_effect_order_code_round_trips`

Fixtures

The fixtures in tests/fixtures/vital/ are generated by cargo xtask vital-fixtures and listed with their SHA-256 in tests/fixtures/PROVENANCE.md. cargo xtask vital-fixtures --check checks them.

File	Holds	Tests
`minimal.vital`	the stamp and the three sample fields	`every_fixture_meets_the_format_contract`
`marks.vital`	every value `preset inspect` marks	the `preset-inspect` command cases
`routes.vital`	three routes, an LFO shape, a two-frame wavetable	`the_routes_fixture_holds_its_generated_values`, `the_routes_wavetable_has_two_full_frames`
`stamps/*.vital`	the version-gate set	`the_stamps_pass_the_version_gate`

The round trip read, write, read is checked for every fixture by every_fixture_meets_the_format_contract.

Limits

Fact	Build	Status	Source
The reader refuses files over 64 MiB and nesting over 128 levels.	this project	verified	unit tests `the_size_cap_applies_to_both_reads`, `deep_nesting_names_the_depth_limit`
Three tables of 257 keyframes come to 8,422,404 bytes of base64. A sample adds more.	source 1.0.6	checked	`wave_source.cpp`
Parsing takes about 3 times the file size for long texts and up to about 38 times for short numbers.	this project	verified	measured in a release build
Ten minutes of fuzzing per target found no crash.	this project	verified	`fuzz/`: 5.4 million reads, 4.0 million round trips, 95 thousand writes

Builds seen

Build	Seen how
source 1.0.6	read, commit `636ca0e`
open engine (`vita` 0.1.0)	probes and load checks of ours
stable 1.0.7	probes and load checks of ours
1.5.3, 1.5.5	sampled presets
early access 1.6.4	reported, not checked

Terms

Factory presets, samples and wavetables are not redistributable, and no file here holds them. The source is GPL-3.0-or-later. This page states facts in our own words and copies no code, comment or table.

Plugin state

A host loads a preset into the VST3 plugin through its state, not through a file. The rows below come from the host spike (issue 244): the retail plugin loaded headless as a library in an x86_64 host process through vst3-host 0.10.0, with no editor window. The probe output is in issue 10.

Fact	Build	Status	Source
The plugin state is the preset JSON plus a `tuning` object, followed by a NUL.	source 1.0.6	checked	`synth_plugin.cpp`
The macOS bundles hold only an x86_64 binary, so a host process must run as x86_64.	1.0.7	verified	`lipo -archs` of the bundle
The plugin loads and renders with no editor window. The probe process wrote nothing to standard error and exited with code 0.	1.0.7	verified	host spike: `minimal.vital` named `SpikeProbe` with `env_1_attack` 1, `filter_1_on` 1 and `filter_1_cutoff` 50, MIDI 60 at velocity 100
The single audio output bus has 2 channels and reports itself inactive after the load. With the bus left inactive, the render was silent; after the host activated it, the render sounded.	1.0.7	verified	host spike: `minimal.vital` named `SpikeProbe` with `env_1_attack` 1, `filter_1_on` 1 and `filter_1_cutoff` 50, MIDI 60 at velocity 100; default patch
The saved component state starts with `VstW`, then `CcnK` at byte 16. The preset JSON starts at byte 176 and is followed by a NUL, zero bytes and the text `JUCEPrivateData`. The JSON holds `tuning` and the app’s stamp `1.0.7`. The plugin saves no controller state.	1.0.7	verified	host spike: `minimal.vital` named `SpikeProbe` with `env_1_attack` 1, `filter_1_on` 1 and `filter_1_cutoff` 50, MIDI 60 at velocity 100; state read-back
A bare preset JSON passed to `IComponent::setState` loads, with or without a trailing NUL. The saved state then holds the name `SpikeProbe` and the key `filter_1_cutoff`.	1.0.7	verified	host spike: `minimal.vital` named `SpikeProbe` with `env_1_attack` 1, `filter_1_on` 1 and `filter_1_cutoff` 50, MIDI 60 at velocity 100; state read-back
A host needs no `VC2!` envelope, the form some Python hosts save: the bare JSON loads.	1.0.7	verified	host spike: `minimal.vital` named `SpikeProbe` with `env_1_attack` 1, `filter_1_on` 1 and `filter_1_cutoff` 50, MIDI 60 at velocity 100
A state without `tuning` keeps the instance’s current tuning: `setStateInformation` loads `tuning` only when the key is there. The app reads no `tuning` from a preset file.	source 1.0.6	checked	`synth_plugin.cpp`, `synth_base.cpp`
`Tuning::jsonToState` reads `scale_start_midi_note`, `reference_midi_note`, `tuning_name`, `mapping_name` and `scale` without a check, and `default` and `mapping` when present. A missing key or a wrong type throws, and the plugin then shows a dialog box. The default tuning starts the scale at MIDI 60 with reference 0 and the scale 0 to 12.	source 1.0.6	checked	`tuning.cpp`, `synth_plugin.cpp`
In one plugin instance, a preset with `reference_midi_note` 2 played MIDI 60 at 293.66 Hz. A preset without `tuning` loaded next still played at 293.66 Hz. With the default tuning added to its state, it played at 261.63 Hz.	1.0.7	verified	`render_vital::a_tuning_does_not_carry_over_to_the_next_preset`: `minimal.vital`, the lead material at 120 BPM
In a host, the plugin sets the engine’s tempo from the host’s tempo on every block, so `beats_per_minute`, in beats per second, follows the host whatever the preset holds. `SoundEngine::setBpm` divides the tempo by 60 in 32-bit floats.	source 1.0.6	checked	`synth_plugin.cpp`, `sound_engine.cpp` `SoundEngine::setBpm`
The plugin stores `f32(bpm) * f32(1 / 60)`, a multiply by the reciprocal: at 140 BPM it saves 2.3333334922790527, and at 100.3 BPM a value two `f32` steps from the correctly rounded 100.3 / 60. A preset with `beats_per_minute` 2 rendered at another tempo saves that value.	1.0.7	verified	`render_vital::a_render_at_another_tempo_passes_the_load_check` at 140 and 100.3 BPM, and `the_saved_tempo_follows_the_model_over_a_sweep`: every whole tempo 4 to 1000, every tenth 4.0 to 400.0 and 600 random tempos on 0.01 and 0.001 steps
A loaded preset renders unlike the default patch: RMS 0.017 for the probe preset against 0.118 for the default patch.	1.0.7	verified	host spike: `minimal.vital` named `SpikeProbe` with `env_1_attack` 1, `filter_1_on` 1 and `filter_1_cutoff` 50, MIDI 60 at velocity 100; default patch
Two renders of one patch differ sample by sample.	1.0.7	verified	host spike: `minimal.vital` named `SpikeProbe` with `env_1_attack` 1, `filter_1_on` 1 and `filter_1_cutoff` 50, MIDI 60 at velocity 100; default patch
In one probe, offline against realtime processing changed the render by less than two repeat renders differ: a difference RMS of 0.0100 against 0.0132.	1.0.7	verified	host spike: `minimal.vital` named `SpikeProbe` with `env_1_attack` 1, `filter_1_on` 1 and `filter_1_cutoff` 50, MIDI 60 at velocity 100
Whether the block size changes the render by more than repeat renders differ. In one probe, blocks of 64 against 512 samples gave a difference RMS of 0.0235, against 0.0132 between two repeat renders. The 64-sample render ran in a second instance.	1.0.7	open	host spike: `minimal.vital` named `SpikeProbe` with `env_1_attack` 1, `filter_1_on` 1 and `filter_1_cutoff` 50, MIDI 60 at velocity 100
`IMidiMapping` returns a parameter for CC 1, CC 7, CC 64, channel aftertouch and pitch bend. A full bend up raised MIDI 60 by 2 semitones.	1.0.7	verified	host spike: default patch, MIDI 60 with and without a full bend; test `a_full_pitch_bend_raises_the_note_two_semitones`
CC 1 from the host moves a route from `mod_wheel`: with the route at -1 on `osc_1_level`, a full wheel lowered the level.	1.0.7	verified	test `the_mod_wheel_reaches_a_route`
A state stamped `1.7.0` loads nothing. The plug-in keeps its init patch and saves the init patch back.	1.0.7	verified	test `a_newer_stamp_falls_back_to_init`
A `settings` key that Vital does not know is gone from the state it saves after the load.	1.0.7	verified	test `a_key_vital_drops_is_a_readback_mismatch`
After a load, the saved state holds every control and the name as sent, a re-encoded `sample.samples` text and a new `version` in each wavetable.	1.0.7	verified	`render` of `routes.vital`
In one probe, a note right after the state load differed from a note after 0.5 s of silence by less than two repeat renders differ.	1.0.7	verified	host spike: `minimal.vital` named `SpikeProbe` with `env_1_attack` 1, `filter_1_on` 1 and `filter_1_cutoff` 50, MIDI 60 at velocity 100
A second instance in the same process, with blocks of 64 samples, loaded the probe preset and rendered it.	1.0.7	verified	host spike: `minimal.vital` named `SpikeProbe` with `env_1_attack` 1, `filter_1_on` 1 and `filter_1_cutoff` 50, MIDI 60 at velocity 100
The standalone app’s `--headless --render` flags.	1.0.7	untestable by rule	the project never runs the app

Open items

Stable 1.0.7 loaded a written preset with its edits, refused the newer stamps and re-saved the preset with every control kept. The ignored test a_preset_saved_by_vital_keeps_the_edits reads that re-save from VITAL_SAVED_PRESET and passes. The unknown-key, no-name, every-source, LFO-shape and wavetable variants loaded in 1.0.7 with no dialog.

A bare preset JSON loads as plugin state in an x86_64 host (see Plugin state).

Still open:

optionally, the stamps and the retail-only keys of early access 1.6.4.
whether the block size changes a render by more than repeat renders differ. The recipe fixes blocks of 512 samples, so renders compare at one block size.

Sources

Vital source, https://github.com/mtytel/vital, GPL-3.0-or-later, facts only, commit 636ca0e.

Keyboard shortcuts

sound-to-patch