Roadmap: alpha and parity estimates; README: video, surround and notation screenshots

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

README.md

@@ -82,6 +82,22 @@ where things are.
     <td><b>Automation and markers.</b> Breakpoint automation for volume, pan, mute, sends and every plugin parameter, written live in Write, Touch and Latch, plus memory locations and a big counter.</td>
     <td><b>Everywhere.</b> Signed builds for macOS (universal), Windows (x64, x86, Arm), Linux (AppImage, deb, rpm, Flatpak), FreeBSD, and a WebAssembly build that runs in the browser.</td>
   </tr>
+  <tr>
+    <td width="50%"><img alt="The Video window showing a colour-bar test movie with a timecode burn-in, above a Video track with thumbnails in the Edit window" src="docs/images/video.png"></td>
+    <td width="50%"><img alt="The Renderer window showing a 7.1.4 speaker layout with two object tracks, beside the Mix window with surround panners and Object/Bed buttons" src="docs/images/surround.png"></td>
+  </tr>
+  <tr>
+    <td><b>Video.</b> A picture track with thumbnails and a Video window that follows the playhead, decoding H.264, ProRes and Motion JPEG in pure Rust, with timecode burn-in and sync offset.</td>
+    <td><b>Surround and immersive.</b> Main and bus formats from stereo to 9.1.6 and Ambisonics, a surround panner with divergence and height, object/bed routing and a Renderer window with live speaker levels.</td>
+  </tr>
+  <tr>
+    <td width="50%"><img alt="The Score Editor showing the lead melody in standard notation on a treble and bass staff" src="docs/images/score.png"></td>
+    <td></td>
+  </tr>
+  <tr>
+    <td><b>Notation.</b> A Score Editor for MIDI tracks, MusicXML export for Sibelius and other notation programs, and printable engraved scores.</td>
+    <td><b>Your plugins too.</b> CLAP, VST3 and Audio Units, with their own editors, their state saved in the session, and every instance created off the audio thread.</td>
+  </tr>
 </table>
 
 ### Feature tour
@@ -191,7 +207,8 @@ and layering checks and the WebAssembly build.
 
 ## Status and roadmap
 
-SoundCraft is **pre-alpha**: the core works and is fun to use, and plenty is still missing. The
+SoundCraft is **pre-alpha, and close to its first alpha**: it covers 94 % of the incumbent's menu
+items and roughly two thirds of its features in depth. The core works and is fun to use, and plenty is still missing. The
 honest status, what's next and our effort estimates are in [`ROADMAP.md`](ROADMAP.md); the
 menu-by-menu comparison is in [`docs/parity.md`](docs/parity.md). Bug reports and wish lists are
 very welcome, in the issues or on Discord.

ROADMAP.md

@@ -8,53 +8,71 @@ the honest status: what works today, what is missing, and how far we are.
 
 | Measure | Value | How |
 |---|---|---|
-| Menu-catalog parity (engine + UI) | **462 / 512 menu items (90 %)**; engine alone 388 / 512 (76 %), see [`docs/parity.md`](docs/parity.md) | `cargo xtask parity` compares the incumbent's 512 menu leaves (names observed black-box) with our command registry; the remaining items are mostly video, Atmos, collaboration, notation and third-party services |
-| Estimated overall feature parity | **~55 %** | judgement across the areas below, weighted by how much professional users rely on them |
-| Estimated remaining effort | **~150–200 wall-clock hours** of Claude Opus 5.5 work (single agent; roughly 50–70 hours with 3–4 agents in parallel) | sum of the per-area estimates below |
+| Menu-catalog parity (engine + UI) | **482 / 512 menu items (94 %)**; engine alone 397 / 512, see [`docs/parity.md`](docs/parity.md) | `cargo xtask parity` compares the incumbent's 512 menu leaves (names observed black-box) with our command registry. Most of what is left is Avid's online services (sign-in, cloud projects, collaboration, Splice, Learn), Dolby Atmos-specific items, HEAT and ARA |
+| Estimated overall feature parity | **~65 %** | judgement across the areas below, weighted by how much working engineers rely on them. Menu items are only the surface; depth (editing feel, plugin catalogue, MIDI and score editing, video) is where the gap is |
+| Distance to an **alpha** | **~85 % there, ~12–18 wall-clock hours** of Claude Opus 5.5 work | see [Alpha](#alpha) below |
+| Distance to **100 % parity** | **~110–150 wall-clock hours** of Claude Opus 5.5 work with a single agent; roughly 40–55 hours with 3–4 agents in parallel (the last three features in this session were built in parallel this way) | sum of the per-area estimates below |
 
 ### Area by area
 
 | Area | Status | Parity | Remaining (h) |
 |---|---|---:|---:|
-| Session model, save/load, undo | Done: tracks, playlists, clips, fades, clip gain, automation, markers, groups, busses; JSON `.scraft` + Audio Files folder; COW undo with gesture coalescing | 80 % | 10 |
-| Audio file I/O | WAV/BWF/RF64, AIFF/AIFC, FLAC (read/write); MP3, OGG, AAC, ALAC, CAF (read); peaks | 75 % | 10 |
-| Editing | Slip/Shuffle/Spot/Grid, cut/copy/paste/clear/duplicate/repeat/shift, separate/heal/trim/consolidate/strip silence, fades & crossfades, nudge, playlists, grabber/trim/selector/smart/pencil/zoom/scrubber gestures, edit groups, playlist lanes with comping, smart-tool fades | 70 % | 25 |
-| Mixer & routing | Faders, pan, mute/solo (SIP, implicit solo), sends pre/post, busses, aux inputs, master faders, VCAs, routing folders, 10 inserts, plugin delay compensation, clip effects | 70 % | 20 |
-| Automation | Breakpoint lanes for volume/pan/mute/sends/plugin params, trim automation; live Write/Touch/Latch passes; write/thin/glide/convert/coalesce commands | 60 % | 18 |
-| Plugins (built-in) | 26 original processors + 2 instruments, AudioSuite offline processing | 45 % | 30 |
-| Third-party plugins (CLAP/VST3/AU) | CLAP and VST3 hosting (audio, params, latency, notes, verified with a real VST3); no plugin GUIs, no state save in sessions, no AU | 55 % | 18 |
-| MIDI | MIDI/instrument tracks, SMF import/export, MIDI editor (piano roll + velocity), event list, step input, quantize/transpose/velocity/duration ops | 40 % | 30 |
-| Recording | Input capture, punch in/out (selection, on the fly, pre/post-roll), loop record into playlists, input monitoring through the channel strip, autosave and recovery | 60 % | 12 |
-| Elastic Audio / TCE / Beat Detective | Pitch-preserving warp on Elastic tracks, TCE to timeline, conform to tempo, Beat Detective and Identify Beat windows | 40 % | 18 |
-| Surround | Main and bus formats from stereo to 9.1.6 and Ambisonics, surround panner with divergence/centre/LFE/height, ITU fold-down, multichannel bounce, meters and device output | 60 % | 15 |
-| Video | Movie audio import; no picture track yet | 10 % | 35 |
-| UI fidelity | Edit + Mix windows, toolbar, rulers, track headers, menus for the whole catalog, floating windows, dialogs | 65 % | 25 |
-| Agent control | CLI, JSON control channel, MCP server (headless + bridged), offscreen UI renders | 90 % (ahead of the incumbent) | 5 |
-| Release engineering | Signed macOS universal, Windows x64/x86, Linux AppImage/deb/rpm/tar/Flatpak, FreeBSD, Web/WASM on every push to `release` | 80 % | 5 |
+| Session model, save/load, undo | Tracks, playlists, clips, fades, clip gain, automation, markers, groups, busses, video; JSON `.scraft` + Audio Files folder; plugin state; autosave; copy-on-write undo with gesture coalescing | 85 % | 8 |
+| Audio file I/O | WAV/BWF/RF64, AIFF/AIFC, FLAC (read/write); MP3, OGG, AAC, ALAC, CAF (read); peaks. Whole files are decoded into memory: disk streaming for very long sessions is still to do | 70 % | 12 |
+| Editing | Slip/Shuffle/Spot/Grid, cut/copy/paste/clear/duplicate/repeat/shift, Paste Special (merge MIDI, markers, clip gain/effects), separate/heal/trim/consolidate/strip silence, fades & crossfades, nudge, playlists with comping, every tool gesture, edit groups, clip groups (and `.scgrp` import/export) | 75 % | 20 |
+| Mixer & routing | Faders, pan, mute/solo (SIP, implicit solo), sends pre/post, busses, aux inputs, master faders, VCAs, routing folders, 10 inserts, plugin delay compensation, clip effects, preamp/instrument/object views | 75 % | 15 |
+| Automation | Breakpoint lanes for volume/pan/mute/sends/plugin params, trim automation; live Write/Touch/Latch passes; write/thin/glide/convert/coalesce. Surround-pan automation is still to do | 60 % | 15 |
+| Plugins (built-in) | 26 original processors + 2 instruments, AudioSuite offline processing for the whole AudioSuite menu | 45 % | 30 |
+| Third-party plugins | CLAP, VST3 and Audio Units (macOS): audio, parameters, latency, notes, state saved in sessions, instances created/destroyed off the audio thread, editors (floating CLAP GUIs, VST3 views on macOS, verified with Melodyne). Missing: AU editors, embedded CLAP GUIs, VST3 editors on Windows/Linux, ARA | 80 % | 10 |
+| MIDI & notation | MIDI/instrument tracks, SMF import/export, MIDI editor (piano roll, velocity, rubber-band select, keyboard transpose/move), event list, step input, quantize/transpose/velocity/duration, real-time properties, Score Editor, MusicXML export (Sibelius) and printable SVG score | 55 % | 20 |
+| Recording | Input capture, punch in/out (selection, on the fly, pre/post-roll), loop record into playlists, input monitoring through the channel strip, autosave and recovery | 60 % | 10 |
+| Elastic Audio / TCE / Beat Detective | Pitch-preserving warp on Elastic tracks, TCE to timeline, conform to tempo, Beat Detective and Identify Beat windows; audio-to-MIDI pitch detection | 45 % | 15 |
+| Surround | Main and bus formats from stereo to 9.1.6 and Ambisonics, surround panner with divergence/centre/LFE/height, ITU fold-down, multichannel bounce, Renderer window, object/bed routing | 60 % | 12 |
+| Video | Picture track with thumbnails, Video and Video Universe windows, H.264 (Baseline/Main/High), ProRes (all flavours) and Motion JPEG decoders in pure Rust, sync offset, relinking. Missing: HEVC/AV1/DNx, timecode-track placement, video export | 55 % | 15 |
+| UI fidelity | Edit + Mix windows, toolbar, rulers, track headers, menus for the whole catalog, floating windows, dialogs (every file command has a path prompt), command palette | 70 % | 20 |
+| Agent control | CLI, JSON control channel, MCP server (headless + bridged), offscreen UI renders | 90 % (ahead of the incumbent) | 3 |
+| Release engineering | Signed macOS universal, Windows x64/x86, Linux AppImage/deb/rpm/tar/Flatpak, FreeBSD, Web/WASM on every push to `release` | 80 % | 3 |
+
+## Alpha
+
+An alpha is a build people can download, open the demo or their own audio in, record, edit, mix
+and bounce without crashes or lost work. Feature-wise SoundCraft is past that bar; what is left is
+shipping and hardening:
+
+1. **Publish the repository and cut the first release** (needs a decision from the maintainers:
+   the GitHub repo, the release secrets, and a push to the `release` branch). The pipelines exist
+   but have never run end to end. ~2 h once unblocked, mostly fixing what the first run finds.
+2. **Platform smoke tests** of the installers on Windows, Linux and FreeBSD (only macOS has been
+   exercised by hand). ~3–4 h.
+3. **Disk streaming** of long sources, so an hour-long multitrack session does not need all of its
+   audio in RAM. ~5–6 h.
+4. **A hands-on QA pass** in the live app: recording on real interfaces, long sessions, plugin-heavy
+   mixes, save/reopen loops. ~3–4 h.
+5. **Docs**: a short user guide next to the existing agent/CLI docs. ~1 h.
 
 ## Robustness
 
 Property tests drive random edit sequences (with full undo), mutated session files and hostile
-command parameters through the engine and mixer; they have already caught fade, overflow and
-unbounded-allocation bugs. Every command also runs with empty parameters on empty and demo sessions.
+command parameters through the engine and mixer; they have caught fade, overflow and
+unbounded-allocation bugs. Every command also runs with empty parameters on empty and demo
+sessions. Hosted plugins live in isolated unsafe crates (`clap-host`, `vst3-host`, `au-host`) and
+are created, loaded and destroyed off the audio thread. The video decoders survive hundreds of
+mutated and truncated movies.
 
 ## Current focus
 
-1. Play back the model-only state the engine already stores: clip effects, trim automation, MIDI real-time properties.
-2. QuickPunch and input monitoring (loop record and Write/Touch/Latch automation passes have landed).
-3. Pitch-preserving realtime Elastic Audio.
-4. Plugin hosting (CLAP first, it is MIT-licensed and Rust-friendly).
-5. Elastic Audio in realtime; Beat Detective workflow window.
-6. README screenshots and the first alpha release.
+1. The alpha list above.
+2. Surround-pan automation; AU editor views.
+3. Built-in plugin depth (more processors, better UIs).
+4. MIDI and score editing depth.
 
 ## Milestones
 
-- **M0–M3 (done):** workspace, gates, model, IO, DSP, mix engine, command engine, Edit/Mix UI.
-- **M4 (done):** realtime playback, metering, plugins, bounce.
-- **M5 (in progress):** editing depth (tools, modes, playlists, fades) — most items landed.
-- **M6 (in progress):** recording.
-- **M7 (in progress):** MIDI editing.
-- **M8 (done for the first cut):** CLI, control channel, MCP, agent acceptance test.
-- **M9:** AudioSuite breadth, Elastic Audio, Beat Detective.
-- **M10:** alpha release on all platforms.
-- **M11:** third-party plugin hosting, video, surround.
+- **M0–M4 (done):** workspace, gates, model, IO, DSP, mix engine, command engine, Edit/Mix UI, realtime playback, metering, plugins, bounce.
+- **M5 (done):** editing depth (tools, modes, playlists, fades, comping, clip groups).
+- **M6 (done for the first cut):** recording, punch, loop record, monitoring.
+- **M7 (in progress):** MIDI and notation editing.
+- **M8 (done):** CLI, control channel, MCP, agent acceptance test.
+- **M9 (in progress):** AudioSuite breadth (done), Elastic Audio, Beat Detective.
+- **M10 (next):** alpha release on all platforms.
+- **M11 (mostly done):** third-party plugin hosting (CLAP, VST3, AU), video, surround.