Two raw materials
Independent sources feed one shared grammar of cutting and recombination. They are material, not decks.
A performance instrument for recorded matter
Cut the source.
Keep the gesture alive.
Navalha 2 turns two sound sources into playable fragments, patterns, gestures and form. It is not a dual deck and not a conventional sampler: the cut itself becomes an instrument.
- JUCEv0.1.0
- REFERENCEPD/web v0.28.1
- STATUSMigration in validation
LIVE SIGNAL MAPENGINE READY
SOURCE AMATTER 01
SOURCE BMATTER 02
MASTER CREATIVE−1.0 dBTPLIVE SAFE
01 / INSTRUMENT
Source → slice → pattern → gesture → form
A sampler whose real material is transformation.
Navalha began as a Pure Data instrument by Glerm Soares. Navalha 2 preserves that lineage while expanding it into a two-source, gesture-driven performance environment directed by Lúcio Araújo. The native JUCE/C++ migration keeps the v0.28.1 PD/web instrument as its functional reference until human acceptance is complete.
Non-destructive slices
Regions, divisions, BLADE and undo reshape access to the source while leaving the original recording intact.
Gesture as memory
TRACE records a movement across BPM and pitch. FORM arranges broader states without flattening live intervention.
Performance, not playback
Patterns, virtual voices and assisted decisions remain reversible and subordinate to PLAY, STOP and REC.
02 / LINEAGE
Original Navalha
Glerm Soares creates the Pure Data instrument and its characteristic slicing/sequencing workflow.
Reference instrument
The interface grows around non-destructive source editing, patterns, gesture, recording and master workflows.
Functional benchmark
The PD/web version remains the explicit baseline for migration and parity decisions.
Native migration
JUCE/C++ v0.1.0 establishes the realtime core, native interface, portable projects and output-safety foundation.
03 / WORKFLOW
One instrument, four workspaces
Prepare matter. Compose behavior. Perform decisions. Keep the takes.
Make the cut playable
- Load SOURCE A/B
- Select and divide regions
- BLADE, WHOLE and UNDO
- Preview without replacing a source
Build a grammar
- GRID, FREE and JITTER clocks
- A/B/GAP step patterns
- FORM scenes and TRACE XY
- Memory and reversible transformations
Act in real time
- Stutter, burst, micro and reverse
- Assisted Performer
- Two virtual voices
- Single or dual-monitor operation
Preserve the work
- REC with five-minute UI limit
- Persistent TAKE Timeline
- Metadata and recipe export
- Track/album tools under validation
05 / QUICK START
A first performance in eight moves
From silence to a recorded take.
- 01
Check audio
Open AUDIO SETUP, choose the output device, sample rate and buffer. Begin with OUTPUT TRIM attenuated.
- 02
Load the sources
Drop WAV material into SOURCE A and B. Use Preview for auditioning without replacing either source.
- 03
Define the region
Select a useful span in the waveform, divide it, then refine with BLADE. Editing remains non-destructive.
- 04
Write a pattern
Place A, B or GAP events in the step grid. Choose GRID, FREE or JITTER as the timing behavior.
- 05
Shape the gesture
Use STUTTER, BURST, MICRO and REVERSE; record an XY TRACE or arrange broader FORM scenes.
- 06
Invite assistance
Enable only the Assisted Performer vocabularies you want. Its edits remain bounded and reversible.
- 07
Perform
PLAY starts the instrument; STOP invalidates pending gestures. MASTER CREATIVE and OUTPUT TRIM serve distinct roles.
- 08
Record and review
Choose PCM16, PCM24 or float32, press REC, then review the finalized WAV in TAKE Timeline.
The native app treats learning as a performance surface, not a help overlay. Ten chapters move from material to take, while the contextual LEARN panel keeps the meaning of a control available at the moment of use — in EN/PT/FR/ES.
04 / TAKES / MASTER
A take is a decision made audible
A take is not residue. It is the work becoming legible.
REC captures the same post-safety signal that reaches the device and the meter: no hidden export path, no substitute mix. A take preserves the moment where cutting, gesture and form become a musical decision that can be heard again, described, compared and carried forward.
Capture without erasing
Each finalized WAV keeps its technical identity and optional metadata. It can return to SOURCE A or B without replacing the original material or concealing the route that produced it.
Review is part of performance
TAKE Timeline turns recording into an editorial space: listen, annotate, rate, compare and retain the recipe of a decision before it becomes memory without context.
Master as a second listening
TRACK and ALBUM tools are not a promise to make everything louder. They prepare a careful second pass: level, sequence, gaps, fades and evidence remain visible before a take becomes an edition.
PERFORM → TAKE → REVIEW → MASTER → WORK / EDITION
06 / INTERFACE
A native shell, not a picture of the web app
The geometry follows the work.
The current JUCE build keeps editing, performance, composition and mix controls visible in one native instrument.
07 / TECHNOLOGY
Native DSP with explicit boundaries
The creative master and the technical output are separate systems.
The application is built in C++20 with JUCE 8. Its realtime path avoids allocation and locks in the audio callback. The current Linux build is verified; Windows and macOS remain planned validation targets.
- Application
- Standalone JUCE/C++
- Current version
- JUCE migration v0.1.0
- Audio sources
- WAV; safe AIFF conversion path
- Recording
- PCM16 / PCM24 with TPDF · float32 clean
- Projects
- Project v2 + portable ZIP
- Verified rates
- 44.1 / 48 / 96 / 192 kHz safety tests
- Live ceiling
- −1 dBTP, stereo-linked lookahead
- License
- GPL-3.0-or-later + JUCE AGPL-3.0-only
01Sources + voicescreative summing
→
02MASTER CREATIVEmusical decision
→
03OUTPUT TRIMtechnical attenuation
→
04LIVE SAFEDC · true peak · lookahead
→
05Device + REC + metersame post-safety signal
Automated evidence: core contracts, golden render, stress, recording soak, project codecs, headless guard and independent FFmpeg true-peak cross-check — 10/10 tests passing on GNU/Linux.
08 / STATUS
Published as work in progress, measured without inflation
Native foundation complete. Final acceptance still open.
- Realtime C++ core and native workspaces
- Two-source slicing and performance grammar
- Audio-thread A/B waveform playhead and time readout
- Project v2, portable project and TAKE catalog
- Live-safe technical output stage
- Deterministic TPDF dither for PCM16/24
- Automated regression and true-peak evidence
- Human parity and long-form performance tests
- Official EBU transient WAV cross-check
- Physical disconnect, reconnect and xrun behavior
- TRACK/ALBUM quality mastering path
- Final clean screenshot gallery
- BS.1770 loudness and post-encoding analysis
- Legacy/quality mastering profiles
- Complete interface localization
- Linux package, then Windows/macOS builds
- Future performance console research
09 / DOCUMENTATION
10 / CREDITS
A lineage kept visible.
Glerm Soares — author, developer and creator of the original Navalha Pure Data instrument.
Lúcio Araújo — conception and direction of Navalha 2 upgrades, musician-facing workflow, interface decisions, testing and musical evaluation.
OpenAI / ChatGPT — implementation assistance, migration tooling, documentation and test preparation under Lúcio Araújo’s direction.
Navalha 2 source code: GPL-3.0-or-later. JUCE 8 is obtained separately under AGPL-3.0-only or a commercial JUCE license. See the license status before distributing binaries.