How We're Redesigning Audacity For The Future

Overview

The video (narrated by Tantacrul) explains how the Audacity team is modernising the app: moving from the messy, decades-old v3 codebase toward a cleaner, Qt-based v4 while balancing existing users’ workflows with making the app intuitive for new users.

Key points:

• Incremental approach (described as “frog boiling”): changes delivered across multiple 3.x releases en route to Audacity 4.
• The audio engine will remain for the first v4 release; a new audio engine and I/O will be introduced later (planned for v5).

“Frog boiling” — deliver changes gradually across 3.x rather than a single disruptive rewrite.

Problems identified

• Large technical debt and an entangled architecture, causing unpredictable breakage and unreliable task estimates.
• wxWidgets UI framework limitations: low visual fidelity, inconsistent cross-platform behavior, limited theming and reduced developer productivity.
• Poor real-time audio: high/unpredictable latency and an outdated audio I/O/engine that hinder synchronized recording and real‑time processing.
• Poor discoverability and confusing modes (six modes in v3), destructive-only effects workflow, and a cluttered UI with too many visible controls.
• Complex plugin ecosystem: VSTs behave inconsistently with many compatibility quirks.

High-level technical decisions

• Replatform the UI from wxWidgets to Qt to enable modern visuals, consistent cross-OS behavior, theming, better performance, and reuse of MuseScore Studio UI components.
• Perform gradual refactors and feature rollouts across 3.x instead of a single rewrite.
• Keep the existing audio engine for v4; replace audio engine and audio I/O in v5 to address latency and real‑time processing needs.
• Invest in automated testing (scripted JS tests driving Audacity) to improve CI, stability and QA coverage.

Feature and product changes — per-release highlights

• Audacity 3.1

  • Clip handles and non-destructive editing (trim without losing source).
  • Groundwork for a clip-centric UX.

• Audacity 3.2

  • Real-time effect support: non-destructive effect racks per track.
  • VST3 support.
  • Major refactor to integrate real-time plugins (took months).

• Audacity 3.3

  • Internal effects converted to real-time for a subset: Bass & Treble, Distortion, Phaser, Reverb, Wahwah.

• Audacity 3.4

  • Musical foundations: snap-to-beats, beat/division units, tempo and time signature support.
  • Integrated a high-quality time-stretch algorithm (from StaffPad; internal open-source integration).

• Audacity 3.5

  • Labels export as subtitle files.
  • Plugin manager and export redesign.
  • Automatic tempo detection (DSP work by Matt; presented at Audio Developer Conference 2024).

• Audacity 3.6

  • Master channel with real-time master effects; compressor and limiter.
  • Intel OpenVINO integration for running local open‑source AI models (e.g., noise reduction).

• Audacity 3.7+

  • Major Qt migration work ongoing.
  • Continued minor releases and bugfixes while porting remaining features.

UX and interaction changes

• Modes reduced significantly (from six to just a couple) by improving direct-manipulation: clip handles and drag-to-move replaced many old modes.
• Robust clip system:
  • Multi-select and grouping.
  • Free pasting/moving without “hitting a brick wall.”
  • Undoable untrim handles.
  • Time-stretch handles.
  • Loop region improvements and export-to-loop behavior that includes trailing effects.
• Sync-lock replaced with explicit grouping plus two ripple-edit behaviors:
  • Track-local ripple: move clips on a single track.
  • Global ripple: move clips on all tracks.
• New split tool (mode) to speed podcast-style edits:
  • Scissor cursor, temporary hold key (S), SHIFT to split across all tracks, drag to cut a hole.
• Delete behavior onboarding:
  • The first time, the app asks whether Delete should leave a gap or close it, then teaches shortcuts.
• Workspaces, customizable top bar and themes (light/dark/contrast and accent colours) let users choose Classic vs Modern layouts and clip appearances (multi-colour or classic).

Plugins and VST support

• Added VST3/AU/LV2 support and real-time plugin stacks.
• Extensive compatibility work required for VSTs due to many third-party idiosyncrasies.

AI features

• Intel OpenVINO integration to run local ML models for features such as noise reduction.
• Pipeline and model support are being refined; caveats remain about model provenance and quality.

Testing, QA and release plan

• Automated test harness using JS scripts to drive Audacity steps and validate results; this catches regressions and frees QA for higher-value testing.
• Alpha build planned (noted as “early November” in the video) with full v4 release targeted early 2026.
  • v4 will not include the new audio engine (audio engine replacement planned for v5).
• Post-v4: expect faster, surgical releases due to reduced technical debt and the Qt migration.

Communications and community handling

• Example: a leaked dev-build video triggered reaction videos claiming Audacity “became a DAW.” The team responded by adding “under construction” messaging, commenting for context, and planning clearer communications.
• Rebrand completed in 2024 (external agency + internal polish). The new icon preserves the headphone motif; Jessica Williamson refined in‑app design work.

The team added “under construction” messaging and context after a leaked dev-build caused amplified/misrepresented reactions.

Developer effort and caveats

• Many features required months of refactoring; VST/plugin support and real-time effect chains were especially time-consuming.
• Some native effects may be delayed in the initial v4 release to ensure stability and parity.
• Possibility of shipping v3 alongside v4 temporarily until parity is reached.

Reviews, external resources and mentions

• MTCHLR — small YouTuber who used a leaked dev build (sparked confusion/misinformation).
• Reaction videos (e.g., Weaver Beats) that amplified or misrepresented the changes.
• Matt’s talk at the Audio Developer Conference 2024 (technical talk on tempo detection / DSP).
• Intel / OpenVINO — partner for local AI models.
• StaffPad team — source of the time-stretch algorithm (open-sourced for integration).

Main speakers and contributors

• Tantacrul — narrator / project lead for the Audacity redesign
• Dominic Mazzoni — Audacity founder (original design intent referenced)
• Core Audacity team:
  • Designers: Alex and Leo
  • Developers: Dmitri, Gabriel, Grzegorz, Matt, Paul
  • Project manager: Yana
  • QA: Sergei and Anton
• MuseScore contributors who helped: Enlur, Igor, Jessica (Jessica Williamson did branding/UI finesse)
• Community contributors acknowledged: Peter Sampson, Steve Daulton, Paul Licameli, David Bailes
• External partners: Intel / OpenVINO, StaffPad
• YouTubers referenced: MTCHLR, Weaver Beats