diff --git a/docs/plans/roadmap-themed-sprints-2026-03.md b/docs/plans/roadmap-themed-sprints-2026-03.md
new file mode 100644
index 00000000..aea9b6c3
--- /dev/null
+++ b/docs/plans/roadmap-themed-sprints-2026-03.md
@@ -0,0 +1,98 @@
+# Research Backlog Themed Sprint Roadmap (#942-#975)
+
+## Scope
+
+- Sources:
+  - `origin/claude/research-daw-synths-KU2At:.llm/research/*`
+  - GitHub issues `#942-#975`
+  - current `main` code synced on 2026-03-27
+- Goal:
+  - convert the research backlog into theme-based sprints
+  - sequence work around shared engine and store dependencies
+  - keep implementation reviewable through small, testable PRs
+
+## Planning Principles
+
+- Ship data model and engine primitives before surface-level UI.
+- Preserve dual-surface parity: every sprint must add or extend `window.__store` actions in addition to visual controls.
+- Prefer PR slices that touch one interaction family at a time and land with browser-testable workflows.
+- Keep dark-theme quality, keyboard access, and agent automation coverage in every sprint, not only at the end.
+- Treat issues `#970`, `#974`, and `#975` as trust multipliers: they improve first-run confidence and should progress in parallel with core audio work.
+
+## Dependency Summary
+
+| Sprint | Theme | Issues | Depends On | Planned PRs |
+|---|---|---|---|---|
+| 01 | Synth Foundation | `#942 #943 #944 #945 #946 #953 #964` | none | 4 |
+| 02 | Sampling, Drum Rack, and Render Pipeline | `#948 #949 #950 #951 #952 #958 #961` | 01 | 4 |
+| 03 | Mixer, Device Graph, and Routing | `#957 #959 #965 #969` | 01 | 4 |
+| 04 | MIDI, Rhythm, and Automation Editors | `#967 #968 #973` | 01, 03 | 3 |
+| 05 | Arrangement and Recording Workflows | `#966 #972` | 02, 04 | 3 |
+| 06 | Onboarding, Library, and Accessibility | `#970 #971 #975` | none | 3 |
+| 07 | Collaboration, Cloud, and Project Lifecycle | `#974` | 06 | 4 |
+| 08 | Advanced Sound Design Lab | `#947 #954 #955 #956 #960 #962 #963` | 01, 02, 03 | 4 |
+
+## Execution Lanes
+
+### Lane A — Audio Platform
+
+- Sprint 01
+- Sprint 02
+- Sprint 03
+- Sprint 08
+
+### Lane B — Editing Workflows
+
+- Sprint 04
+- Sprint 05
+
+### Lane C — Experience and Trust
+
+- Sprint 06
+- Sprint 07
+
+## Recommended Order
+
+1. Land Sprint 01 first to replace preset-only instrument assumptions with a real instrument state model.
+2. Start Sprint 06 in parallel once Sprint 01 is underway, because onboarding/accessibility mostly avoids audio-engine conflict.
+3. Run Sprint 02 and Sprint 03 after Sprint 01 because both need richer instrument and effect schemas.
+4. Start Sprint 04 once Sprint 03 has automation-arm and device parameter hooks available.
+5. Start Sprint 05 after Sprint 02 and Sprint 04 so recording, comping, and clip operations share the new render and editor primitives.
+6. Start Sprint 07 after Sprint 06 so first-run and project-library UX do not diverge from cloud and share-player flows.
+7. Keep Sprint 08 last because it depends on the synth, sampler, routing, and modulation foundations from Sprints 01-03.
+
+## Branch and PR Conventions
+
+- Sprint branches:
+  - `feat/v0.0.x-sprint-01-synth-foundation`
+  - `feat/v0.0.x-sprint-02-sampling-drum-render`
+  - `feat/v0.0.x-sprint-03-mixer-device-routing`
+  - `feat/v0.0.x-sprint-04-midi-rhythm-automation`
+  - `feat/v0.0.x-sprint-05-arrangement-recording`
+  - `feat/v0.0.x-sprint-06-onboarding-library-accessibility`
+  - `feat/v0.0.x-sprint-07-collaboration-cloud`
+  - `feat/v0.0.x-sprint-08-advanced-sound-design`
+- PR naming:
+  - `feat: sprint 01 PR1 instrument state model`
+  - `feat: sprint 03 PR2 device chain UX`
+  - `feat: sprint 06 PR1 accessible control primitives`
+
+## Done Criteria For Every Sprint
+
+- `npx tsc --noEmit`
+- `npm run build`
+- targeted Vitest coverage for the new engine/store surface
+- at least one browser workflow covering human interaction and one `window.__store` workflow covering agent interaction
+- no new console noise outside explicit error handlers
+- updated docs/plans for any scope change discovered during implementation
+
+## Sprint Documents
+
+- [Sprint 01 — Synth Foundation](./sprint-01-synth-foundation.md)
+- [Sprint 02 — Sampling, Drum Rack, and Render Pipeline](./sprint-02-sampling-drum-render.md)
+- [Sprint 03 — Mixer, Device Graph, and Routing](./sprint-03-mixer-device-routing.md)
+- [Sprint 04 — MIDI, Rhythm, and Automation Editors](./sprint-04-midi-rhythm-automation.md)
+- [Sprint 05 — Arrangement and Recording Workflows](./sprint-05-arrangement-recording.md)
+- [Sprint 06 — Onboarding, Library, and Accessibility](./sprint-06-onboarding-library-accessibility.md)
+- [Sprint 07 — Collaboration, Cloud, and Project Lifecycle](./sprint-07-collaboration-cloud.md)
+- [Sprint 08 — Advanced Sound Design Lab](./sprint-08-advanced-sound-design.md)
diff --git a/docs/plans/sprint-01-synth-foundation.md b/docs/plans/sprint-01-synth-foundation.md
new file mode 100644
index 00000000..c1363cdc
--- /dev/null
+++ b/docs/plans/sprint-01-synth-foundation.md
@@ -0,0 +1,86 @@
+# Sprint 01 — Synth Foundation
+
+## User Stories
+
+- As a producer, I want every instrument track to expose editable synth parameters, so that I can shape sounds instead of swapping only preset names.
+- As a producer, I want subtractive and FM synth building blocks to behave like real DAW instruments, so that ACE-Step can cover bread-and-butter bass, lead, pad, and keys workflows.
+- As an AI agent, I want synth state to be represented in the project store instead of hidden inside Tone instances, so that I can automate instrument design through `window.__store`.
+
+## Problem
+
+- Issues `#942 #943 #944 #945 #946 #953 #964` all describe expected baseline synth behavior, but the current app still treats instruments as a preset string plus a fixed Tone setup.
+- The backlog mixes UI asks with engine asks, yet the real blocker is missing instrument state and routing primitives.
+
+## Root Cause
+
+- `src/types/project.ts:7-13` models instruments as `TrackType` plus a `SynthPreset` string, which cannot represent oscillator, filter, modulation, or engine-specific state.
+- `src/engine/SynthEngine.ts:10-52` hardcodes six preset branches on `Tone.PolySynth(Tone.Synth)` and never persists parameter snapshots.
+- `src/engine/SynthEngine.ts:66-139` caches one synth per track keyed only by preset, so changing a single parameter would currently require replacing the whole instance.
+- `src/components/pianoroll/PianoRoll.tsx:122-155` and `src/components/pianoroll/PianoRoll.tsx:224-260` expose sampler config and editor toolbar state, but there is no synth editor surface or store-backed synth parameter UI.
+
+## Solution
+
+### Deliverables
+
+- Introduce a track-level `instrument` model that separates engine kind from preset label.
+- Add a subtractive synth schema with oscillator, amp envelope, filter, filter envelope, LFO, and unison settings.
+- Add an FM synth schema with at least carrier/modulator ratio, level, waveform, and modulation index.
+- Build a synth editor panel with visual envelope sections and collapsible advanced controls.
+- Add preset snapshots backed by factory JSON plus IndexedDB user presets.
+
+### Issue Map
+
+- `#942` synth parameter editing baseline
+- `#943` filter envelope
+- `#944` routeable LFO
+- `#945` unison/detune
+- `#946` FM instrument
+- `#953` preset browser/save/load
+- `#964` visual synth editor
+
+### Proposed PR Slices
+
+1. `feat: add track instrument state model and migration`
+   - extend `src/types/project.ts`
+   - add store migration/helpers in `src/store/projectStore.ts`
+   - expose agent-safe actions in `src/main.tsx`
+2. `feat: add subtractive synth parameter engine and editor essentials`
+   - update `src/engine/SynthEngine.ts`
+   - add synth editor component under `src/components/pianoroll/`
+   - support `#942` and `#943`
+3. `feat: add LFO, unison, and FM engine support`
+   - implement `#944 #945 #946`
+   - add visual status/routing previews
+4. `feat: add preset browser and user preset persistence`
+   - implement `#953 #964`
+   - ship factory preset JSON and IndexedDB user presets
+
+## Verification
+
+- `npx tsc --noEmit`
+- `npm run build`
+- `npx vitest run src/engine/__tests__/SynthEngine.test.ts src/store/__tests__/projectStore.test.ts`
+- add E2E coverage:
+  - create a piano-roll track
+  - change oscillator, filter, envelope, and preset in UI
+  - verify the same state through `window.__store.getState().project`
+- manual workflow:
+  - create Bass patch
+  - duplicate track
+  - switch duplicate to FM Bell
+  - save and reload project
+  - confirm both instruments restore correctly
+
+## Files To Touch
+
+- `docs/plans/sprint-01-synth-foundation.md`
+- `src/types/project.ts`
+- `src/store/projectStore.ts`
+- `src/engine/SynthEngine.ts`
+- `src/components/pianoroll/PianoRoll.tsx`
+- `src/components/pianoroll/`
+- `src/main.tsx`
+- `src/services/` preset persistence helpers
+- `tests/e2e/`
+- `src/engine/__tests__/`
+- `src/store/__tests__/`
diff --git a/docs/plans/sprint-02-sampling-drum-render.md b/docs/plans/sprint-02-sampling-drum-render.md
new file mode 100644
index 00000000..071ab5b9
--- /dev/null
+++ b/docs/plans/sprint-02-sampling-drum-render.md
@@ -0,0 +1,80 @@
+# Sprint 02 — Sampling, Drum Rack, and Render Pipeline
+
+## User Stories
+
+- As a producer, I want sampler instruments to use realistic zones, slices, and warp behavior, so that browser-based playback does not feel like a toy pitch shifter.
+- As a beat-maker, I want the drum machine to accept my own samples and pad-level shaping, so that I can build real kits instead of only choosing four synthesized kits.
+- As a producer and AI agent, I want freeze and bounce workflows to reduce CPU while preserving reversibility, so that large projects stay stable.
+
+## Problem
+
+- Issues `#948 #949 #950 #951 #952 #958 #961` all sit on the same missing platform: the app has one-sample playback, synthesized drum pads, and partial bounce primitives, but not a general sample-instrument architecture.
+
+## Root Cause
+
+- `src/types/project.ts:57-81` only supports a single `SamplerConfig` buffer with trim and ADSR, and `src/types/project.ts:96-108` only gives each drum pad `sampleKey`, `volume`, and `pan`.
+- `src/engine/SamplerEngine.ts:81-220` manages one `AudioBuffer` per track and derives pitch purely from `playbackRate`, which blocks multi-zone, velocity-layer, slice, and pitch-independent stretch workflows.
+- `src/engine/DrumEngine.ts:46-220` constructs every drum sound from fixed Tone synth factories instead of per-pad instrument descriptors and effect chains.
+- `src/store/projectStore.ts:2292-2385` exposes `freezeTrack`, `unfreezeTrack`, `flattenTrack`, and `bounceInPlace`, but the state model and UX stop short of a complete freeze/unfreeze workflow.
+- `src/components/sequencer/DrumMachineEditor.tsx:170-260` only exposes kit selection and per-pad volume details, not sample loading, tuning, or pad processing.
+
+## Solution
+
+### Deliverables
+
+- Replace the single-buffer sampler model with zone-based sampler definitions.
+- Add velocity layers, round-robin, and slice maps that remain scriptable through the store.
+- Add per-pad sample mode, tune, envelope, filter, drive, and send state to the drum machine.
+- Add pitch-independent time stretch and warp markers for sampler clips and slices.
+- Complete track freeze/unfreeze/bounce UX on top of the existing offline render hooks.
+
+### Issue Map
+
+- `#948` multi-sample zones
+- `#949` user sample loading on pads
+- `#950` per-pad effects and tuning
+- `#951` velocity layers and crossfading
+- `#952` audio slice mode
+- `#958` warp/time-stretch
+- `#961` freeze/bounce track to audio
+
+### Proposed PR Slices
+
+1. `feat: add sampler zone model and engine migration`
+   - zone schema, velocity ranges, round-robin
+2. `feat: add drum pad sample mode and per-pad shaping`
+   - sample loading, choke groups, pad filter/drive/send
+3. `feat: add slice mode and warp-ready sample UI`
+   - transient slicing, manual slices, warp marker model
+4. `feat: add freeze-unfreeze workflow and render progress`
+   - full `#961` UX, store actions, browser tests
+
+## Verification
+
+- `npx tsc --noEmit`
+- `npm run build`
+- `npx vitest run src/services/__tests__/bounceInPlace.test.ts src/store/__tests__/drumMachine.test.ts`
+- add new unit coverage:
+  - sampler zone selection by pitch and velocity
+  - slice-to-MIDI mapping
+  - freeze/unfreeze persistence
+- browser workflows:
+  - drag a sample onto a drum pad, trigger it, reload project
+  - load a sample into sampler, create slices, trigger them from piano roll
+  - freeze a synth track, confirm CPU-saving audio playback, unfreeze it
+
+## Files To Touch
+
+- `docs/plans/sprint-02-sampling-drum-render.md`
+- `src/types/project.ts`
+- `src/store/projectStore.ts`
+- `src/engine/SamplerEngine.ts`
+- `src/engine/DrumEngine.ts`
+- `src/components/pianoroll/QuickSamplerEditor.tsx`
+- `src/components/sequencer/DrumMachineEditor.tsx`
+- `src/components/pianoroll/`
+- `src/hooks/useTransport.ts`
+- `src/services/bounceInPlace.ts`
+- `src/services/audioFileManager.ts`
+- `tests/e2e/`
+- `src/store/__tests__/`
diff --git a/docs/plans/sprint-03-mixer-device-routing.md b/docs/plans/sprint-03-mixer-device-routing.md
new file mode 100644
index 00000000..8456ff0d
--- /dev/null
+++ b/docs/plans/sprint-03-mixer-device-routing.md
@@ -0,0 +1,75 @@
+# Sprint 03 — Mixer, Device Graph, and Routing
+
+## User Stories
+
+- As a mixer, I want channel strips and device chains to expose real routing, metering, and insert behavior, so that I can mix inside ACE-Step without treating the mixer as a placeholder.
+- As a producer, I want routing-heavy effects such as convolution reverb and sidechain ducking, so that modern mix workflows are possible.
+- As an AI agent, I want device parameters and routing choices to exist in store state, so that automated mix flows can modify them safely.
+
+## Problem
+
+- Issues `#957 #959 #965 #969` look separate, but they are all symptoms of one limitation: there is no unified device graph between the effect model, mixer surface, and audio routing layer.
+
+## Root Cause
+
+- `src/components/mixer/MixerPanel.tsx:12-18` caps inserts and sends at four and two, and `src/components/mixer/MixerPanel.tsx:144-247` renders fixed sections without I/O routing, dB markings, or advanced metering.
+- `src/components/mixer/EffectChain.tsx:51-141` encodes effect presets inline and `src/components/mixer/EffectChain.tsx:171-279` renders a basic device card row without rack nesting, A/B states, preset library, or searchable add flow.
+- `src/engine/EffectsEngine.ts:97-227` builds linear Tone node chains, ships only algorithmic reverb, and has no first-class effect-rack or sidechain input abstraction.
+- `src/engine/EffectsEngine.ts:233-247` rebuilds a single linear chain per track, which is insufficient for parallel racks, convolution branches, and visual routing feedback.
+
+## Solution
+
+### Deliverables
+
+- Introduce a track device graph abstraction that supports serial and parallel chains.
+- Upgrade channel strips with RMS plus peak metering, dB scale, insert management, pre/post sends, solo-safe, and simple I/O routing.
+- Rebuild the device chain UX with searchable add, collapse, reorder, preset persistence, and rack containers.
+- Add convolution reverb and sidechain routing as first-class graph nodes.
+
+### Issue Map
+
+- `#957` convolution reverb
+- `#959` sidechain routing UI
+- `#965` pro-grade mixer panel
+- `#969` device-view redesign
+
+### Proposed PR Slices
+
+1. `feat: add mixer meter upgrade and channel-strip state`
+   - RMS/peak meters, dB scale, insert slot expansion
+2. `feat: add searchable device chain UX and drag reorder`
+   - searchable add menu, collapse, preset surface
+3. `feat: add convolution reverb and visual processor feedback`
+   - IR loading, EQ curve, gain reduction meter
+4. `feat: add sidechain routing and send-state upgrades`
+   - sidechain source selector, pre/post send toggle, solo-safe
+
+## Verification
+
+- `npx tsc --noEmit`
+- `npm run build`
+- `npx vitest run src/engine/__tests__/effectsEngineNativeNode.test.ts src/store/__tests__/sendsReturns.test.ts tests/unit/StatusBar.test.tsx`
+- add new unit coverage:
+  - device graph serialization
+  - sidechain routing selection
+  - convolution preset persistence
+- browser workflows:
+  - create a reverb rack and reorder devices
+  - set kick as sidechain source for bass compressor
+  - verify insert count, send toggles, and meter behavior in mixer
+
+## Files To Touch
+
+- `docs/plans/sprint-03-mixer-device-routing.md`
+- `src/types/project.ts`
+- `src/store/projectStore.ts`
+- `src/engine/EffectsEngine.ts`
+- `src/components/mixer/MixerPanel.tsx`
+- `src/components/mixer/EffectChain.tsx`
+- `src/components/mixer/EffectCards.tsx`
+- `src/components/mixer/LevelMeter.tsx`
+- `src/components/ui/Knob.tsx`
+- `src/components/mixer/VerticalFader.tsx`
+- `src/engine/__tests__/`
+- `src/store/__tests__/`
+- `tests/e2e/`
diff --git a/docs/plans/sprint-04-midi-rhythm-automation.md b/docs/plans/sprint-04-midi-rhythm-automation.md
new file mode 100644
index 00000000..3d55f73e
--- /dev/null
+++ b/docs/plans/sprint-04-midi-rhythm-automation.md
@@ -0,0 +1,68 @@
+# Sprint 04 — MIDI, Rhythm, and Automation Editors
+
+## User Stories
+
+- As a composer, I want piano-roll editing tools that feel musical instead of merely technical, so that note editing stays fast.
+- As a beat-maker, I want Elektron-style step detail and probability controls, so that the drum workflow moves beyond an on/off grid.
+- As a producer, I want automation to support recording, drawing, and modulation tools, so that parameter movement is part of composition instead of an afterthought.
+
+## Problem
+
+- Issues `#967 #968 #973` are editor issues, but the shared blocker is that note, step, and automation data are all modeled too simply for advanced interactions.
+
+## Root Cause
+
+- `src/components/pianoroll/PianoRoll.tsx:27-120` keeps ghost-note visibility and editor mode state locally and does not derive scale-aware behavior from project harmony.
+- `src/components/pianoroll/PianoRollCanvas.tsx:178-238` supports note creation and chord stamping, but the drag model is single-note focused and lacks native operations for humanize, legato, strum, and multi-note stretch.
+- `src/components/pianoroll/VelocityLane.tsx:19-99` only draws velocity bars; it does not own an interactive editing model for painted velocity curves.
+- `src/engine/DrumEngine.ts:11-27` keeps drum steps to `active` plus `velocity`, which cannot represent probability, micro-timing, ratchets, or per-step parameter locks.
+- `src/components/timeline/AutomationLaneView.tsx:37-165` renders a simple polyline editor with draggable points but no recording modes, tool palette, curve handles, or automation-arm state.
+
+## Solution
+
+### Deliverables
+
+- Extend piano-roll actions with scale highlight, lock-to-scale, velocity painting, quantize strength, legato, humanize, strum, and multi-note stretch.
+- Extend step data with probability, micro-timing, pitch offset, ratchet, and step-length metadata.
+- Add automation editor tool state, recording-arm state, curve handles, LFO tool generation, and copy/paste flows.
+
+### Issue Map
+
+- `#967` piano-roll UX
+- `#968` drum machine step UX
+- `#973` automation recording and toolset
+
+### Proposed PR Slices
+
+1. `feat: add piano-roll scale, velocity, and transform tools`
+2. `feat: add drum step metadata and Elektron-style editing`
+3. `feat: add automation arm, draw tools, and LFO generation`
+
+## Verification
+
+- `npx tsc --noEmit`
+- `npm run build`
+- `npx vitest run src/store/__tests__/drumMachine.test.ts src/store/__tests__/projectStore.test.ts`
+- add new unit coverage:
+  - quantize-strength math
+  - legato and humanize transforms
+  - step probability and micro-timing evaluation
+  - automation curve and LFO tool generation
+- browser workflows:
+  - create chord notes, apply strum, paint velocity, legato them
+  - build a drum pattern with step probability and ratchets
+  - arm a filter parameter and record automation while transport plays
+
+## Files To Touch
+
+- `docs/plans/sprint-04-midi-rhythm-automation.md`
+- `src/types/project.ts`
+- `src/store/projectStore.ts`
+- `src/components/pianoroll/PianoRoll.tsx`
+- `src/components/pianoroll/PianoRollCanvas.tsx`
+- `src/components/pianoroll/VelocityLane.tsx`
+- `src/components/sequencer/`
+- `src/engine/DrumEngine.ts`
+- `src/components/timeline/AutomationLaneView.tsx`
+- `src/store/__tests__/`
+- `tests/e2e/`
diff --git a/docs/plans/sprint-05-arrangement-recording.md b/docs/plans/sprint-05-arrangement-recording.md
new file mode 100644
index 00000000..300f3385
--- /dev/null
+++ b/docs/plans/sprint-05-arrangement-recording.md
@@ -0,0 +1,65 @@
+# Sprint 05 — Arrangement and Recording Workflows
+
+## User Stories
+
+- As a producer, I want clip operations and take comping directly on the arrangement, so that editing audio feels like a real DAW workflow.
+- As a vocalist or instrumentalist, I want punch ranges, visual count-in, latency feedback, and comp lanes, so that recording sessions are trustworthy.
+- As an AI agent, I want arrangement editing and record-state transitions to be explicit store actions, so that end-to-end browser tests can validate them without canvas-only gestures.
+
+## Problem
+
+- Issues `#966` and `#972` both depend on a more deliberate arrangement interaction model. The code already contains pieces of take-lane and punch logic, but the user-facing flow is incomplete.
+
+## Root Cause
+
+- `src/components/timeline/Timeline.tsx:208-260` focuses on viewport, selection, and drag windows rather than clip-operation commands such as split, duplicate, loop, reverse, and comp actions.
+- `src/hooks/useRecording.ts:64-117` and `src/hooks/useRecording.ts:119-201` already accumulate loop and punch takes, but the workflow ends at saving clips and toggling take-lane visibility.
+- `src/engine/RecordingEngine.ts:25-169` captures device, monitoring, and level-meter state, but does not expose latency reporting, pre-roll, tuner state, or input clip warnings as a dedicated recording surface.
+- `src/store/projectStore.ts:2292-2385` contains freeze/flatten/bounce primitives that can support arrangement clip ops, yet clip-operation shortcuts and comp editing are not organized around those primitives.
+
+## Solution
+
+### Deliverables
+
+- Add arrangement clip commands for split, duplicate, loop, reverse, fade handles, and crossfade behavior.
+- Build a recording panel or transport extensions for punch-in/out, pre-roll, latency display, count-in visuals, tuner, and input clip warnings.
+- Complete take-lane selection and comp editing on arrangement tracks.
+
+### Issue Map
+
+- `#966` timeline clip operations
+- `#972` recording UX
+
+### Proposed PR Slices
+
+1. `feat: add arrangement clip operations and shortcuts`
+2. `feat: add punch, pre-roll, latency, and visual count-in`
+3. `feat: add take-lane comp editing and recording polish`
+
+## Verification
+
+- `npx tsc --noEmit`
+- `npm run build`
+- `npx vitest run src/store/__tests__/toggleClipMuted.test.ts src/store/__tests__/splitClipAtZeroCrossing.test.ts`
+- add new unit coverage:
+  - clip split/duplicate/loop actions
+  - punch-range clip placement
+  - take comp selection
+- browser workflows:
+  - split and duplicate clips from keyboard shortcuts
+  - record with punch-in/out and visual count-in
+  - create loop takes and comp the best parts on take lanes
+
+## Files To Touch
+
+- `docs/plans/sprint-05-arrangement-recording.md`
+- `src/store/projectStore.ts`
+- `src/hooks/useRecording.ts`
+- `src/engine/RecordingEngine.ts`
+- `src/components/timeline/Timeline.tsx`
+- `src/components/timeline/TrackLane.tsx`
+- `src/components/timeline/ClipBlock.tsx`
+- `src/components/layout/Toolbar.tsx`
+- `src/components/tracks/TrackHeader.tsx`
+- `src/store/__tests__/`
+- `tests/e2e/`
diff --git a/docs/plans/sprint-06-onboarding-library-accessibility.md b/docs/plans/sprint-06-onboarding-library-accessibility.md
new file mode 100644
index 00000000..b266f4f3
--- /dev/null
+++ b/docs/plans/sprint-06-onboarding-library-accessibility.md
@@ -0,0 +1,71 @@
+# Sprint 06 — Onboarding, Library, and Accessibility
+
+## User Stories
+
+- As a new user, I want the first-run path to get me to music in under 30 seconds, so that ACE-Step feels AI-native instead of empty.
+- As a creator, I want a modern sample browser with preview, search, and drag-to-timeline, so that discovery stays inside the app.
+- As a keyboard or assistive-tech user, I want the DAW surface to be navigable and legible, so that accessibility is built into the core experience.
+
+## Problem
+
+- Issues `#970 #971 #975` all affect first impression and trust. Today ACE-Step has useful starter pieces, but they are fragmented and do not yet form a guided, accessible creation flow.
+
+## Root Cause
+
+- `src/components/dialogs/NewProjectDialog.tsx:89-152` and `src/components/dialogs/NewProjectDialog.tsx:207-245` provide templates and demo starters, but there is no first-run path selector, progress tracking, or contextual help system.
+- `src/components/assets/LoopBrowser.tsx:74-191` supports categories, search, preview, and drag payloads, but it is still a loop library rather than a full sample browser with BPM/key metadata, favorites collections, recent items, or AI sample generation.
+- `src/components/ui/Knob.tsx:137-210` and `src/components/mixer/VerticalFader.tsx:110-127` expose partial ARIA but omit `aria-valuetext`, value announcements, and shared accessibility primitives.
+- `rg` coverage shows almost no reduced-motion, high-contrast, color-blind, skip-link, or focus-trap infrastructure beyond isolated `aria-live` regions.
+
+## Solution
+
+### Deliverables
+
+- Build a first-run shell with three entry paths: generate, template, blank.
+- Promote Loop Browser into a sample browser service with metadata, favorites, recents, preview sync, and AI sample generation entry points.
+- Standardize accessible control primitives, skip links, reduced-motion handling, and dialog focus management.
+
+### Issue Map
+
+- `#970` first-run onboarding
+- `#971` sample browser
+- `#975` accessibility
+
+### Proposed PR Slices
+
+1. `feat: add accessible control primitives and navigation affordances`
+   - slider ARIA, skip links, reduced motion, focus management
+2. `feat: add sample browser metadata, preview sync, and collections`
+   - extend current loop library into a real browser
+3. `feat: add AI-first onboarding and contextual quick help`
+   - welcome dialog, template path, help panel, shortcut overlay
+
+## Verification
+
+- `npx tsc --noEmit`
+- `npm run build`
+- `npx vitest run tests/unit/statusBarControls.test.ts tests/unit/StatusBar.test.tsx`
+- add new unit coverage:
+  - accessible slider attributes
+  - sample browser filtering and favorites
+  - first-run state gating
+- browser workflows:
+  - first launch from clean storage
+  - choose "Generate a Song", complete quick-start
+  - preview and drag a sample to timeline
+  - complete the same key actions with keyboard-only navigation
+
+## Files To Touch
+
+- `docs/plans/sprint-06-onboarding-library-accessibility.md`
+- `src/components/dialogs/NewProjectDialog.tsx`
+- `src/components/assets/LoopBrowser.tsx`
+- `src/components/assets/LoopBrowserItems.tsx`
+- `src/components/ui/Knob.tsx`
+- `src/components/mixer/VerticalFader.tsx`
+- `src/components/ui/DualRangeSlider.tsx`
+- `src/store/uiStore.ts`
+- `src/index.css`
+- `src/App.tsx`
+- `tests/e2e/`
+- `tests/unit/`
diff --git a/docs/plans/sprint-07-collaboration-cloud.md b/docs/plans/sprint-07-collaboration-cloud.md
new file mode 100644
index 00000000..7d5a5d17
--- /dev/null
+++ b/docs/plans/sprint-07-collaboration-cloud.md
@@ -0,0 +1,67 @@
+# Sprint 07 — Collaboration, Cloud, and Project Lifecycle
+
+## User Stories
+
+- As a creator, I want projects to persist in the cloud with version history and share links, so that my work survives beyond one browser instance.
+- As a collaborator, I want shared playback and later real-time sync to build on the same project model, so that collaboration does not fork into a separate product.
+- As an AI agent, I want cloud and share actions to be explicit APIs, so that automated flows can publish, restore, and validate remote project state.
+
+## Problem
+
+- Issue `#974` is larger than one feature. It combines storage, sharing, project organization, version history, MIDI IO, and collaboration foundations. Shipping it safely requires a dedicated lifecycle sprint rather than scattered UI additions.
+
+## Root Cause
+
+- `src/services/cloudStorageService.ts:61-190` stores cloud projects in memory and shared projects in `localStorage`, which makes sharing local-only and non-persistent across environments.
+- `src/store/collaborationStore.ts:12-74` tracks viewer mode, share token, and collaborator metadata, but has no backend session model, sync transport, or conflict state.
+- `src/services/projectStorage.ts:14-119` is optimized for local IndexedDB project storage, not remote version history, folder organization, or shared ownership.
+- `src/main.tsx:122-133` exposes sharing helpers for the local share API, but not a remote repository client or cloud action surface.
+
+## Solution
+
+### Deliverables
+
+- Introduce a project repository abstraction with local and remote implementations.
+- Add persistent remote save/load/list/version APIs and wire them into the collaboration store.
+- Build a version timeline, hosted share-player load path, project organization metadata, and MIDI import/export integration.
+- Define the real-time sync foundation as phase two on top of the same repository and session model.
+
+### Issue Map
+
+- `#974` collaboration and cloud workflow foundation
+
+### Proposed PR Slices
+
+1. `feat: add project repository abstraction and cloud API client`
+2. `feat: add cloud save-load-version history UI`
+3. `feat: add hosted share player and project organization metadata`
+4. `feat: add collaboration session foundation for cursors and track locks`
+
+## Verification
+
+- `npx tsc --noEmit`
+- `npm run build`
+- `npx vitest run src/services/__tests__/aceStepApi.test.ts src/store/__tests__/projectStore.test.ts`
+- add new unit coverage:
+  - remote repository adapter behavior
+  - version-history restore logic
+  - share-token load flows
+- browser workflows:
+  - save project to cloud, refresh, restore it
+  - publish a share page and open it in a clean session
+  - restore an earlier version from the timeline
+
+## Files To Touch
+
+- `docs/plans/sprint-07-collaboration-cloud.md`
+- `src/services/cloudStorageService.ts`
+- `src/store/collaborationStore.ts`
+- `src/services/projectStorage.ts`
+- `src/services/projectSharingService.ts`
+- `src/components/sharing/`
+- `src/components/dialogs/ShareDialog.tsx`
+- `src/components/dialogs/ProjectListDialog.tsx`
+- `src/main.tsx`
+- `src/store/projectStore.ts`
+- `tests/e2e/`
+- `src/services/__tests__/`
diff --git a/docs/plans/sprint-08-advanced-sound-design.md b/docs/plans/sprint-08-advanced-sound-design.md
new file mode 100644
index 00000000..2c626760
--- /dev/null
+++ b/docs/plans/sprint-08-advanced-sound-design.md
@@ -0,0 +1,71 @@
+# Sprint 08 — Advanced Sound Design Lab
+
+## User Stories
+
+- As a sound designer, I want wavetable, granular, physical, additive, and spectral engines, so that ACE-Step can move beyond “browser DAW” expectations and become distinctive.
+- As an expressive performer, I want MPE and a modulation matrix, so that per-note motion and controller input can shape advanced engines naturally.
+- As an AI agent, I want advanced sound engines to reuse the same instrument and modulation model from earlier sprints, so that automation and preset generation remain coherent.
+
+## Problem
+
+- Issues `#947 #954 #955 #956 #960 #962 #963` are the most differentiated features in the backlog, but they should not land until the core synth, sampler, and routing layers are stable.
+
+## Root Cause
+
+- `src/engine/SynthEngine.ts:10-139` currently assumes `Tone.Synth` or simple preset playback and has no engine plug-in boundary for wavetable or physical modeling voices.
+- `src/types/project.ts:9-18` and `src/types/project.ts:113-124` lack engine-specific instrument schemas and per-note expression metadata.
+- `src/engine/EffectsEngine.ts:97-247` handles simple Tone node chains, but there is no AudioWorklet DSP layer for granular or spectral processors.
+- `src/main.tsx:22-92` exposes store state to agents, yet there is no advanced modulation bus or per-note expression surface for those agents to target.
+
+## Solution
+
+### Deliverables
+
+- Add a pluggable instrument-engine boundary on top of Sprint 01’s track instrument state.
+- Add a modulation matrix and per-note expression model that can be shared by wavetable and MPE-enabled engines.
+- Implement advanced DSP engines in an AudioWorklet-backed layer where real-time FFT or grain scheduling is required.
+
+### Issue Map
+
+- `#947` wavetable
+- `#954` granular
+- `#955` physical modeling
+- `#956` modulation matrix
+- `#960` MPE
+- `#962` additive
+- `#963` spectral processing
+
+### Proposed PR Slices
+
+1. `feat: add modulation matrix and wavetable engine`
+2. `feat: add granular and physical-modeling instruments`
+3. `feat: add additive and spectral processor surfaces`
+4. `feat: add MPE and per-note expression UI`
+
+## Verification
+
+- `npx tsc --noEmit`
+- `npm run build`
+- add new unit coverage:
+  - modulation routing math
+  - wavetable frame selection
+  - MPE channel allocation
+  - spectral processor serialization
+- browser workflows:
+  - load wavetable patch and modulate position with LFO
+  - play granular instrument from a loaded sample
+  - record MPE gestures and verify per-note expression state
+
+## Files To Touch
+
+- `docs/plans/sprint-08-advanced-sound-design.md`
+- `src/types/project.ts`
+- `src/store/projectStore.ts`
+- `src/engine/SynthEngine.ts`
+- `src/engine/`
+- `src/components/pianoroll/`
+- `src/components/mixer/`
+- `src/main.tsx`
+- `src/services/`
+- `src/engine/__tests__/`
+- `tests/e2e/`