Vital Connection Reference

Recovered visual constructions, 2026-09-20. Read alongside the working knowledge (a separate local project record, not included in this Atlas publication), not as a replacement for it. The offline visual atlas contains 152 article figures and 71 PDF pages, each with its own finding and preserved source. The complete findings retain individual examples, controls and distinctions beyond this navigation map.

These are usable documented mechanisms and visibly demonstrated configurations. They need not pass a new experiment before they may inform construction. This review does not create new native measurements or establish perceptual success. Where an existing source or native observation refines a diagram's wording, retain that specific correction, not a blanket rejection of the document.

Reading A Construction

A useful representation has three interacting parts:

A table is an ingredient, not this whole representation. Conversely, a complex table is not intrinsically wrong: retain it where it contributes, and use native operations where they explain the required behavior more effectively. These are available roles for any reference, not six hard-coded sound categories.

1. Independent Audio Branches

Oscillators and sampler can select Filter1, Filter2, both, Effects or Direct Out. Filter source buttons and oscillator destination controls describe the same routing. Distinguish a source's power from its audible level and destination. For example, one component can retain its body while another receives high-pass shaping; a third can bypass voice filters. Filters need not be one final EQ on a premixed resource. Direct Out also bypasses the effects path.

Use this when components require different shaping or articulation. Preserve each branch's role and the point of summation; do not automatically duplicate tables into L/R oscillators. Sources: M05, M07, V14-V15, T09, F19, F22.

2. Independent Amplitude And Spectral Envelopes

ENV1 contributes amplitude and voice lifetime. A separate envelope can move a filter cutoff, resonance, oscillator morph or ingredient level. The T19 example explicitly connects ENV2 to Filter1 cutoff. F17 coordinates cutoff and resonance from one LFO with separate amounts. A shared gesture does not require identical amounts, base values or remaps at its destinations.

Retain delay, attack, hold, decay, sustain level, release and segment curvature. Release can interrupt earlier stages; a still envelope drawing is not a complete performance. This enables a short bright gesture over a longer body, or slower spectral motion over a quick amplitude onset, without encoding both into frames. Sources: T19, M18, V41, F17-F18, F23.

3. Filter Shape, Excitation And Tracking

The eight families are Analog, Dirty, Ladder, Digital, Diode, Formant, Comb and Phaser. Family/subtype, blend, cutoff, resonance, drive, mix and keytracking are distinct choices. The six V36 plots retain low-pass, high-pass, band-pass, notch, double-notch and peak behavior. Movement can traverse a response shape as well as move its cutoff; changing drive can change excitation and nonlinear response.

The Diode Low Shelf/Low Cut comparison retains a concrete difference in low-body retention while blend moves toward band-pass. Partial dry/wet combination has phase-dependent consequences. Keytracking controls how the spectral structure relates to the played register. These are alternatives to repeatedly changing harmonic amplitudes in a table, not mandatory additions to every patch. Sources: V35, V36-V37, M13-M15, T15.

4. Coordinated Resonances

Formant AOIE exposes X/Y vowel structure, transpose, Peak and Spread. Comb distinguishes resonance-spacing pitch from its additional Cut control. Positive and negative phaser-filter responses differ. A moving multi-peak/notch structure is not equivalent to moving a low-pass or changing amplitude alone.

These families provide editable resonant structures when that representation helps. A listener's vowel-like description does not uniquely diagnose a Formant filter; ordinary resonant filtering or other modulation can also contribute. Sources: V38, V39-V40, M15.

5. Carrier, Modulator And Modulation Envelope

The FM article's sequence enables two sine sources, reduces the modulator's audible level, selects FM OSC2 on OSC1, then connects ENV2 to OSC1 FM amount. The modulator still has a role when not independently audible. Its tuning, phase, resource and depth trajectory influence the carrier. Powering it off is a different operation from reducing its audible level.

F16 changes the carrier resource while keeping a simple modulator. F20 chains OSC3 -> OSC2 -> OSC1; T41 further illustrates Sample -> OSC3 -> OSC2 -> OSC1. F15 retains RM from oscillator/sample as a different interaction. These are dedicated audio dependencies, not a claim that any matrix route runs at audio rate. The finite oscillator budget must include modulator roles, not only summed layers. Sources: F10, F11-F16, F20, F24-F25, T41.

6. Two Independent Oscillator Morph Stages

The spectral stage includes Vocode, Formant Scale, harmonic/inharmonic stretch, Smear, random amplitudes, LP/HP, phase dispersion, Shepard Tone and spectral time skew. The wave stage includes Sync, Formant, Quantize, Bend, Squeeze, Pulse and FM/RM dependencies. Frame selection, these morph stages and unison coexist.

For example, T21 modulates wave Formant, not a Formant filter or spectral Formant Scale. T35 separately controls oscillator morph amount and phase. Preserve exact parameter identity and where the transformation occurs. Simple resources can support rich evolving output through these operations. Sources: V14, M09-M10, T11, T21, T35-T36.

7. Modulation Of Modulation

The matrix offers per-connection enable, polarity, stereo, Morph, amount and remap. Remap is a function of source value, not an independent clock. A macro can target another connection's Amount; a Random can target LFO frequency; macros can target envelope timing and other macros. One connection can therefore determine how another acts, rather than merely adding a second independent trajectory.

Retrieve assignments through source highlighting, target contribution display and matrix rows. Retain bypassed routes as bypassed, not absent. The two states in C24 must not be merged into one invented graph. Existing native evidence also retains velocity-controlled filter-envelope depth under its tested conditions. Sources: V10, V08-V11, C24-C25, T35-T37; native block (separate local experimental record: vital_live_learning_20260920.md; not published here).

8. Motion Shape Versus Playback Rules

LFO geometry, start/loop position, smoothing, delay, rate domain and channel phase are separate from its mode. Trigger, Sync, Envelope, Sustain Envelope, Loop Point and Loop Hold retain different timing/release behavior; the complete descriptions remain in M19-M20 and the saved chapter. Rate domains include seconds, tempo, dotted/triplet and keytrack. Preserve actual points and curve powers, not only a shape name that may remain unchanged after editing.

Choose a note-relative gesture, held behavior or ongoing motion as appropriate to the construction. A resource frame axis does not supply those rules by itself. Source: V42, M19-M20, T20, F21, F26.

9. Structured Variation And Performance

Random styles include Perlin, Sample & Hold, Sine Interpolate and Lorenz, with rate, reset and channel behavior. Note, Velocity, Lift, Oct Note, Pressure, Slide, Stereo and per-note Random have distinct meanings. T48 coordinates three oscillator transposes from one Sample & Hold source with snap; T51 maps velocity to cutoff; T50 maps Note to reverb mix. These are different relationships from global velocity-to-amplitude tracking.

Use appropriate input domains for register, strike, release and expression. Preserve mono/poly target aggregation instead of assuming an FX slot becomes an independent effect for every note. Existing source-backed Stereo and Lift semantics remain in the working notebook. Sources: V12, V13, M21-M23, C23, T22, T47-T52.

10. Sample As An Independent Ingredient

Sample pitch/keytracking affects playback speed and duration; random start, looping and alternating direction have separate effects. A repeatable onset and a varying texture need different settings. The sample can enter its own filter branch or serve an FM/RM dependency. F27 maps LFO1 to White Noise level and high-passes that branch separately from the carrier.

Retain the actual resource and playback controls alongside its level envelope. This is not a granular/time-stretch engine, and a complex residual is not automatically noise. Sources: V15, M10-M12, T12-T14, F15, F27.

11. Unison As Distributed Synthesis States

Count, detune amount, distribution/power, Advanced range, blend, starting phase, phase randomization and stereo spread describe more than width. Stacks can include center-drop, octaves, chords and harmonics. Table, spectral and distortion spread distribute different synthesis states among voices. Zero detune does not imply a single identical voice when phases or other state spreads differ.

These are distinct from independently recorded L/R tables or chorus delay modulation. Assign each mechanism a job; do not use two panned sources by default or outlaw them when independently evolving content actually warrants them. Sources: V57, V14, M08, M40-M41, T10, T38; native unison/chorus observations (separate local experimental record: vital_live_learning_20260920.md; not published here).

12. Dynamics As A Time- And Level-Dependent Stage

The compressor supports multiband/low/high/single layouts, upper and lower threshold/ratio regions, band gains, attack, release and mix. Graph edges and inner bars are interactive threshold and ratio controls, not decoration. V48-V49 retain explicit static upward/downward compression examples. An upper 1:1 region does not bypass lower processing or band gain.

Input level and detector history matter. Band dynamics can alter articulation and balance differently from static EQ or a note envelope. This stage can also change how hard a later distortion is excited. Preserve the operating level instead of normalizing away every gain dependency during construction. Sources: V45, V46-V49, M29-M30, T27; native timing/level observations (separate local experimental record: vital_live_learning_20260920.md; not published here).

13. Nonlinearity And Its Surrounding Filters

Soft/Hard Clip, Linear/Sine Fold, Bit Crush and Down Sample are distinct modes. Drive, mix and internal None/Pre/Post filtering shape different relationships. Pre filtering changes excitation; post filtering shapes generated content. Compression -> distortion and distortion -> compression are different graphs. Likewise a voice filter before summation and an FX filter after summation need not create the same response under polyphony.

F28-F29 preserve compression -> pre-filtered soft clipping -> EQ -> FX-filter steps in one article sequence, not a mandatory universal chain. This is compact native timbre generation from simpler ingredients, with level and motion intact. Sources: V51, V43, F28-F29, T25, T29; native filter/distortion order observations (separate local experimental record: vital_live_learning_20260920.md; not published here).

14. Static Balance And Moving Final Shaping

EQ retains three independently configured bands with gain, cutoff, resonance and shape, even when only one tab is visible. The FX Filter retains the broader filter-family vocabulary at a post-voice location. Either can receive modulation where supported; they are not necessarily static finishing touches.

Use a final response for shared balance and voice branches for component-specific behavior. Keep stage location and motion explicit; do not absorb every spectral correction into extracted harmonics. Sources: V52, V53, M34, T30-T31, F29.

15. Chorus, Flanger And Phaser Relationships

Chorus combines multiple modulated delays, two delay groups, depth, feedback, spectral shaping and mix. Flanger combines delay-derived coloration with center, depth, feedback, offset and rate. Phaser has a different response and additional blend control. Freeze allows static coloration instead of compulsory motion.

These can contribute moving spectral structure and channel differences, not only stereo width. Retain source state separately from delayed/filtered motion. Feedback sign and dry/wet combination are consequential, not automatically beneficial. Sources: V44, V54-V55, M35-M37, T26, T32-T33.

16. Echo And Tail Structure

Delay provides Mono, Stereo, Ping Pong and Mid Ping Pong, two time controls, feedback, filtering and mix. Reverb separates pre-low/high cuts, low/high shaping in its feedback structure, internal chorus, pre-delay, size, time and mix. Retain the source-backed EQ placement from the notebook rather than inferring an external post-reverb equalizer from a diagram label.

Echo timing, release envelope and reverberant decay are different contributors. Modulation can make their amount depend on note/expression, but the sum/history behavior matters. Reverb/delay are available tools, not obligatory solutions for every residual or missing width. Sources: V50, V56, M31, M38, T28, T34, T50.

17. Resource Authoring And Extraction Choices

Wave, Line and Audio File Sources have their own keyframes; source/modifier lanes compose a resource rather than adding runtime oscillators. Magnitude and phase bars are separate. Blend may be waveform, spectral, smooth or absent. A copied LFO contour can become a cycle without bringing its original timing rules.

Audio import distinguishes Wavetable, Vocode and Pitch Splice. The retained example shows 2048-sample fixed import versus a detected 336.2 window for the other two, with different phase treatment. These are example settings, not a universal analysis rate. Preserve extraction parameters and the resource so extraction and execution failures remain distinguishable.

Phase Shift, Wave Window, Frequency Filter, Slew Limiter, Wave Folder and Wave Warp provide additional authored transformations. Within-cycle windowing is not ADSR; editor harmonic-coordinate filtering is not runtime absolute-Hz filtering; complex phase mixing can change magnitudes. Keep editable recipes where useful, alongside native runtime controls. Sources: V16, V17-V34, C26-C27, T56.

18. Inter-Note State And Faithful Execution

Polyphony, note priority, Kill/Steal, glide amount/slope, Always Glide, Octave Scale and Legato govern transitions and overlap. Oversampling and Hi-Res Wavetable are independent quality settings. The UI waveform/spectrum display is useful live feedback, not a recorded time-frequency measurement or a full per-voice trace.

Saving a complete preset, saving a wavetable and synthesizing a preset into a table preserve different state. Native export/reload must retain the intended resources and connections; existing sample-save caveats remain in the notebook. None of this review changed host settings, source presets or experiment results. Sources: V02, V03-V07, V17, V57, M39-M43, T39, T53-T57.

Applying The Record

Retrieve a relevant construction before replacing an existing useful component. Its source IDs lead to the exact image, explanation, original document and hashes. Read neighboring steps when the mechanism is a sequence. Preserve observed gains at their demonstrated scope; neither a failed aggregate match nor absence of a previous test erases an available mechanism.

When a future change succeeds or fails, add the actual configuration, context and result at the existing evidence location and link it from the relevant role. The useful outcome is a growing, retrievable set of constructions and conditions, not repeated document ingestion or a promise that conversation memory is permanent.