MGA Sonic Anvil Manual

1. What this app does

2. Quick start

1. Drop a Wave / AIFF / MP3, or press Ctrl+O (multiple files open as tabs; the status bar shows progress).
2. Listen, cut, fade, and place markers or a loop on the waveform. For multichannel work, enable layouts on Settings → Shown first, then pick one with SPEAKER on the status bar. Record with Ctrl+R.
3. Save with Ctrl+S as Wave or MP3 (AIFF export is not supported). Ctrl+Shift+M is MP3-only. Writing an MP3 is like Save As: it writes a file and keeps the current tab (the written MP3 is not opened). MP3 is re-encoded from 16-bit PCM. A valid LAME path in Settings is used; empty or invalid falls back to Windows (default 192 kbps). Markers are not written.
4. To send it to Wwise, turn on WAAPI above the level meter, pick a destination, and press EXPORT.

3. Layout

• Settings — gear at the top left, or Ctrl+Shift+O. Tabs are General / Shown / Audio / Editing / Export. Shown is a checklist of speaker definitions to enable (Stereo only by default; the definition in use stays listed until you switch away; at least one must stay enabled; the status-bar SPEAKER menu uses the same list). General is UI language (Auto / Japanese / English; Auto is Japanese if the OS is Japanese, otherwise English) and file associations (Wave / AIFF / MP3; check to make this app the default, uncheck to remove; already-default types are checked; writes to this exe immediately). Audio starts with a speaker layout (mono through Atmos, including game-oriented variants such as Quad, LCRS, 2.1, and 5.1 Side; you do not add your own; the file itself stays numbered; each speaker is assigned a file channel (Ch), and only those lanes show L / C / LFE and so on; the default is 1, 2, 3…; switching layouts without OK asks whether to save the current one). Inside it are the API (WaveOut / WASAPI / ASIO), the audio device, record / playback ports and Ch (speaker names come from the layout, port names from the device; Ch is shared by record and playback; Off leaves the speaker unused; record has level meters), and per-playback-port Sine −20 dB (1 kHz, or 80 Hz on LFE) plus an English spoken channel-name Voice (not on LFE). Stereo playback with no port map still downmixes the speakers. Editing is the loudness target (LKFS) and default fade curves. Export is MP3 (Windows bit rate when not using LAME / LAME path / LAME options) and Parallel export count (default Auto = a quarter of the CPU cores; the dropdown max is half the cores). Width follows the Audio tab. Ctrl+Tab goes to the next tab, Ctrl+Shift+Tab to the previous. Esc closes it.
• Overview — top strip. The bright area is the current view. Drag to move (scrubs the center); wheel zooms around the playhead.
• Waveform — the edit surface: time ruler, marker / region flags, playhead, mouse guide. Triple-click selects all. Click a channel name on the left to solo (click the same name again to clear; Ctrl+click adds; Shift+click mutes / unmutes). Tab solos a channel (again for the next; Shift+Tab goes backward; past the last clears it; the time selection does not change; fade / volume / normalize / reverse / pitch / time stretch / copy / delete apply only to that channel; muted lanes go gray; works while playing). A slow load shows Now Loading ... in the center.
• Time scrollbar — under the waveform. Pan the current view.
• Tabs — shown even for a single file. Close with Ctrl+W. Quit with Ctrl+Q. Reopen a closed tab with Ctrl+Shift+T (more than one). Right-click for “Close Other Tabs (O)”, “Close This and Tabs to the Right (R)”, “Close This and Tabs to the Left (L)”, “Close All Tabs (W)”, “Select All Tabs (A)”, “Export Wave (E)”, “Export MP3 (M)”, “Export Wave Separated by Markers (K)”, and “Export Wave Separated by Regions (G)”. Export writes the current edits to another file (dirty tabs stay dirty). One tab opens a Save dialog; several tabs ask for a folder. Overwrite is confirmed when the destination already exists: Yes = overwrite, No = skip those files, Cancel = stop (a long list shows the count and the first names only). Separate export always asks for a folder. Markers need two or more spans; regions need at least one (tabs that do not qualify are skipped). Names append _1 (or _01 and so on so the digits match the count). Batch names use the source file name (or the tab title if unsaved); a clash gets 2 and up. The last folder is remembered. While it runs, a frosted-glass overlay blocks the UI and shows an overall bar plus a bar for every song encoding in parallel (two columns when they would not fit). Ctrl+Shift+W also closes all tabs. Closing several tabs with two or more dirty ones also offers “Save all and quit” and “Quit without saving any”. Ctrl+Shift+Alt+M exports every tab as MP3 (keeps the current tabs; there is no key for batch Wave). Select tabs with Ctrl+click (toggle), Shift+click (range), or Ctrl+A while the pointer is over the tab bar (all); selected tabs are highlighted and Esc deselects. With a selection, the right-click menu becomes “Close Selected/All Tabs” / “Paste Copied Edits to Selected/All Tabs” / Wave or MP3 export / separate by markers or regions, and Ctrl+V applies the copied edit history to every selected tab at once. When tabs no longer fit, inactive tabs shrink first. Scroll arrows appear only if a title would fall below 6 characters.
• Transport — top row is playback and zoom; second row is fade / normalize / delete / save / MP3 output, Tips on/off, and manual ?. License / GitHub sit on the bottom of that row.
• Always on Top — right end of the status bar. Keeps the window in front.
• Spectrum / loudness / meters / vector scope — a view of the playback output. Loudness sits left of the spectrum (Short Term / Integrated / Momentary Max, Loudness Range, True Peak; target LKFS is in Settings; values are blue = headroom, orange = approaching, red = over; holds the last reading after stop and measures again on a new play). Left of loudness is the edit-history list (as many rows as fit; older rows clip; click opens edit history with U; the strip itself does not select). The spectrum mixes every channel into the same LED-band display as Layer Music Checker. Waveform channel names come from the speaker-to-Ch map (L / C / LFE and so on on assigned lanes; unassigned lanes stay numbered from 1; the default is 1, 2, 3…). For three or more channels, names are white on a colored square in a common appearance order (16 colors for Atmos 9.1.6). The waveform lanes take a light wash of the same colors. The level-meter bars use the same colors and do not show names. Stereo and mono use the cyan gradient: the inner two bars are Peak (red lamps clip), the outer two are RMS (both have hold lines). Numbers below are Peak then RMS. Three or more channels replace the vector scope with a surround view that uses the same map (file headers are not used to invent names). Energy uses the same channel colors as the waveform. Size is channel level; the outline texture is that channel's recent waveform (the tip is now, the flanks are a few milliseconds earlier). One thin dark-gray outline wraps the overall spread. The outline holds peaks briefly, then falls slowly. The level meter shows one Peak bar per channel, with Peak hold and a slower RMS hold. Drag the left edge of the column to widen it (the default is the minimum; it stops when the bars reach full thickness; the width is remembered on the next launch). After stop, the display slowly fades. WAAPI above the level meter toggles the bar.
• Tips — above the WAAPI bar, fixed height. Hover a control to see its description; overflow scrolls. The first tip appears immediately; moving to another control or away waits 300 ms before replacing or clearing. Click the heading or body to lock (cyan frame, Tips (Locked)); click again to unlock. The text does not change while locked or while the pointer is on the panel. Transport buttons also show a short tooltip (with the shortcut; wraps, no ellipsis). Other controls have no floating tooltip.
• WAAPI bar — connection, destination, Play -E, Auto Active, EXPORT.
• Status bar — format info (name is on the tab; sample rate / bit depth / channels turn red after conversion until you save; size is MB (MiB / B), and turns red as an estimate after conversion or a range delete that changes length); SPEAKER left of NOW POS switches the speaker layout (same list as Shown; the definition in use stays listed until you switch away); current / selection start / selection length / selection end / total (all but total are editable; wheel / ↑↓ nudge — time by 1 s / Shift 10 s / Ctrl 1 min / Ctrl+Shift 10 min, samples by 1 / Shift 10 / Ctrl 100 / Ctrl+Shift 1000; right-click switches time and samples); Always on Top on the right.

4. Playback

• Space — play / stop. Stop returns to the start position.
• Enter — pause in place.
• Ctrl+Space — from 3 seconds earlier.
• Alt+Enter — restart from the playback start.
• Ctrl+drag — scrub. Every channel is downmixed to L/R (the level meter moves on L/R only).
• G — type NOW POS in the status bar. Enter jumps. Copy / paste allowed. Right-click switches time / samples.
• Z or . (the numpad . is the same) — center. Toggles center-lock while playing; clears it when stopped.
• L — loop from 3 seconds before the end of the selection (or the sample loop / -L).
• Tab — solo a channel. Other lanes go gray and stay muted. Again moves to the next channel; Shift+Tab goes backward. Past the last clears solo. Click a channel name on the left to solo (click the same name again to clear; Ctrl+click adds; Shift+click mutes / unmutes). The time selection does not change. Waveform edits apply only to that channel. Works while playing or stopped.
• Ctrl+R — record into a new tab using the speaker layout and Ch map. Press again, or Space / Enter / Esc, to stop. Record ports, playback ports, and Ch are assigned inside that layout (Ch is shared by record and playback; Off leaves the speaker unused).
• Stereo playback with no port map still downmixes the speakers to L/R. Mono with two or more ports plays the same signal on L and R.

5. Selection and editing

Drag to select. Double-click selects a span (between markers). Triple-click always selects all. Esc or a move without Shift clears it. Fade / volume / pitch / normalize / reverse / time stretch with nothing selected uses the whole file in time. While a channel is soloed, those edits apply only to the audible lanes. You can also type the selection start / length / end in the status bar.

• FADE IN I / FADE OUT O — pick a curve, Space to preview, Enter to apply. 1–9 select a curve.
• X — linear fade around the playhead in the view (before = out / after = in).
• V — shows the loudness curve and opens volume. Type a dB value, or use ↑↓ / wheel by 0.1 dB (Shift 1 / Ctrl 3 / Ctrl+Shift 6). Space previews, Enter applies. The menu shows how whole-file Integrated LKFS / RMS / Peak will change. The Short Term curve follows the same dB. Shift+V closes the curve.
• P — pitch. ↑↓ / wheel by a semitone (Shift or Ctrl for an octave). Space previews, Enter applies. Range is ±2 octaves. Keep length is on by default (Tab moves to the checkbox). Off also changes length, like playback speed. A frosted-glass overlay shows progress while it runs.
• T — time stretch (pitch stays). Uses the selection, or the whole file. Destination time and ratio (%) stay linked. ↑↓ / wheel (time: 1 s / Shift 10 s / Ctrl 1 min / Ctrl+Shift 10 min; ratio: 0.1 / Shift 1 / Ctrl 10 / Ctrl+Shift 25%). Tab switches fields. Space previews, Enter applies. Range is 10–1000%. A frosted-glass overlay shows progress while it runs.
• NORMALIZE N — fit the peak to -0.1 dB.
• R — reverse. Uses the selection, or the whole file. Audio only; markers / regions / loops stay put.
• DELETE — ripple-delete the selection. Selected markers / regions go with it.
• Ctrl+X / C / V — cut / copy / paste (markers in range included; a matching region is included too).
• U — edit history. ↑↓ move, Enter apply, Esc cancel. Clicking the list left of loudness also opens it.
• History copy & paste — inside the history, Ctrl+click to select steps (Shift+click or Shift+↑↓ for a range, Ctrl+A for all), then Ctrl+C. Press Ctrl+V in another file to re-apply the same processing (its history window does not need to be open). The copy persists until you copy something else, so you can switch tabs and press Ctrl+V repeatedly to apply it to multiple files. Ctrl+C with no selection copies every step up to the current position. Ranges are mapped by time across sample rates, and gain-type edits such as normalize are recomputed on the target.
• Ctrl+Z / Ctrl+Y (Ctrl+Shift+Z also redo) — undo / redo.

6. Markers / regions / loops

• M / Ins — marker at the playhead when there is no selection. The same frame cannot take another. With a selection, the first press places start and end (or the midpoint if both ends already have markers); further presses replace them with thirds, then fourths, and so on (hold uses the same repeat timing as marker nudge; one history entry until you release).
• Shift+L — set the selection as the sample loop (same range again clears it).
• Shift+R — set the selection as a region. If both ends already have a region, the first press divides it in two; the same range again divides it in three, and so on (clear with right-click / Delete; hold uses the same repeat timing, one history entry until you release).
• Click a flag to select an edge. Shift+click for a range, Ctrl+click to add. Drag or ←/→ to move (Shift ×3).
• Double-click a comment / region name. -A lime, -L blue, -E red, -R gray. See 9-3 for what EXPORT does with them.
• Ctrl+Shift+R — rename.
• Right-click, or Ctrl+Del, deletes markers.

7. Spectrogram / loudness analysis

A cycles spectrogram / overlay. Shift+A returns to the waveform. V shows loudness analysis (the Short Term curve) and opens volume input. Shift+V returns to the waveform. The VIEW loudness button toggles the view only. L loops from 3 seconds before the end of the selection (or the sample loop / -L). Loudness analysis draws the Short Term LKFS curve. The white width is ±1 LU around the curve (1 LU = 1 LKFS; the same “approaching” window as the meter), drawn as a translucent fill. The waveform is a Mid-mix peak-dBFS line on the same vertical axis as LKFS (0 dBFS = 0 on the scale). At sample-level zoom it connects each sample with a line, same as the regular waveform. dBFS and LKFS are not the same, but peak and loudness can be compared at the same height. The background bands use the same blue / orange / red as the meter (red from 0 to the target, orange for the next 6 LU, blue below). The curve is primary cyan / orange / red; the red segment is a little thicker. The left scale is LKFS (waveform dB in the regular view). The white line is the Settings target. Overlay draws the waveform shape on top of the spectrogram (no translucent wash). It does not paint -A / -L / -E / -R, sample-loop, or region fills (flags and lines stay; spectrogram-only and loudness analysis also skip those fills). The overlay waveform is also a single Mid-mix of the channels. At sample-level zoom it uses the same dots and connecting lines as the regular waveform. At amplitude zoom 1 it uses the middle half; amplifying fills the full height. Colors are display-normalized so the file's peak reads as 0 dBFS (the audio itself is untouched); the gain is measured once at load / edit analysis time, never in real time. Dark bins are lifted so they stay readable, without crushing the highlights. In spectrogram / overlay, drag the bar on the left or hold Alt+↑ / ↓ to change the lift. The frequency axis is logarithmic and tighter toward the bass (below 100 Hz is barely visible). The top of the scale is 24 kHz. When the whole file fits on screen — such as a short SE at fit — each screen column runs its own FFT and draws sharp, without interpolation. Any zoomed view draws smoothly interpolated columns from the full-file cache, keeping zooming and follow-scroll light. Scrolling shifts columns and computes only what is missing.

8. Format conversion

• S — sample rate
• B — bit depth
• C — file channel count (Mono / Stereo; this is not the speaker layout)

Presets are 8 / 11.025 / 16 / 22.05 / 32 / 44.1 / 48 / 96 kHz, bit depths 4 / 8 / 16 / 24, and channels Mono / Stereo. Custom Hz is 1000–384000, nudged with ↑↓ / wheel by 1 Hz (Shift 10 / Ctrl 100 / Ctrl+Shift 1000). The menu sits just off the cursor. Highlighting a value updates the status size to the estimate. A frosted-glass overlay blocks the UI while converting. Confirmed fields stay red until you save. Sample rate and bit depth always convert from the PCM captured at load (or after a content edit such as fade). 48 kHz → 24 kHz then 11 kHz is rebuilt from 48 kHz, not cascaded through 24 kHz. Space previews rate and bit depth only (channels apply with Enter; 1–2 pick a row). 4-bit is for quantized audition. Wave writes only 8 / 16 / 24, so a 4-bit document is written as 16-bit.

9. WAAPI EXPORT (Wave-only)

The same assembly as Wwise IM Importer’s Wave-only mode. Keep one-shot and loop material on a single master waveform instead of pre-separating files. EXPORT writes that wave as one file into Originals and maps ranges with MusicClip trim. It creates one Music Playlist Container. No Custom Cues. There is no bar / tempo overlay, no multi-wave Switch, and no XML marker-track mode.

This app is also a waveform editor, so fades and other edits are baked into the source before it is handed over. Markers and the sample loop are written into the WAV on save.

9-1. Steps

1. Start Wwise Authoring.
2. Turn on WAAPI above the level meter (W).
3. Select a destination object in Wwise. Keep Target can pin the path.
4. The export folder must be under Originals of the connected project.
5. EXPORT or Ctrl+Shift+E. Enabled only while WAAPI is connected.

• Play -E (E) — play after the Exit Cue. In-app preview double-plays -E on a loop wrap so two playheads appear. EXPORT writes Play post-exit on the Music Playlist Container default transition rule (Any to Any).
• Auto Active — bring Wwise to the front after EXPORT.

9-2. How ranges are built

Earlier items win.

1. Sample loop (the WAV smpl chunk, or Shift+L) — before the loop is Intro, the loop is -L, after it is -E. This beats markers.
2. Suffixed markers (-A / -L / -E, or a name containing Loop) — each starts a region that runs to the next split. A bare span right after -L is given -E automatically.
3. Exactly two unsuffixed markers — the span between them becomes an implicit loop.
4. Regions (Shift+R) — used as-is when there is no sample loop and no special markers.
5. If none of the above apply, the whole file is one range.

9-3. Suffixes

Markers are used for preview and splitting only. They are not written as Wwise Custom Cues.

9-4. What Wwise gets

• One Music Playlist Container (named after the file). Sequential play; each range is a Playlist Item.
• Music Segments. One range keeps the file name; several become _a / _b…. Loop ranges use Loop Count 0 (infinite). -A and -E share a Segment with their neighbor; only the Entry / Exit Cues move.
• The source wave is not split. Every clip references the same WAV and uses Begin / End Trim for the range.
• The first Segment is Zero Latency, 50 ms look-ahead, 500 ms prefetch. Later Segments use 500 ms look-ahead.
• Tempo 120 / 4/4 is overridden (there is no bar management).
• With Play -E on, Play post-exit is set on the playlist default rule (Any to Any).

10. Shortcuts

This section lists every implemented shortcut, including ones already described earlier. Contextual keys apply only while that UI is open. The waveform Tips panel shows the same operations.

File / tabs

Playback

View / zoom

Seek / selection

Edit

Markers / regions / loops

Format

WAAPI

Mouse / pointer

Contextual

While edit history is open (other shortcuts do not apply)

While picking a fade curve

While the volume menu is open

While the pitch menu is open

While the time-stretch menu is open

While the format-conversion menu is open

While editing a status-bar time field

In Settings, Esc closes the window. The UI ignores input while a format conversion, tab export, or multi-file open is running.

11. Settings and language

• Settings are on the gear at the top left, or Ctrl+Shift+O, split into General / Shown / Audio / Editing / Export. Shown is a checklist of speaker definitions to enable (Stereo only by default; the definition in use stays listed until you switch away; at least one must stay enabled; the status-bar SPEAKER menu uses the same list). File associations on General are per extension (.wav / .wave / .aif / .aiff / .mp3). Check to make this app the default; uncheck to remove it. Types already using this app start checked. The change is written for the current user to the running exe and does not wait for OK. Audio starts with a speaker layout (mono through Atmos; you do not add your own). Inside it are the API (WaveOut / WASAPI / ASIO), the device, record / playback ports and Ch (shared by record and playback; Off leaves the speaker unused; default 1, 2, 3…; switching layouts without OK asks whether to save the current one), and per-playback-port Sine −20 dB (1 kHz, or 80 Hz on LFE) plus an English spoken channel-name Voice (not on LFE). Stereo playback with no port map still downmixes the speakers. The record meters confirm a signal. Loudness target (default −24 LKFS, from −70 to 0; color scale only, does not change the audio) and default fade curves are on Editing; MP3 and Parallel export count are on Export. The settings window remembers its position and height when closed. Width is set automatically from the Audio tab. Ctrl+Tab goes to the next tab, Ctrl+Shift+Tab to the previous.
• MP3 uses the user-supplied lame.exe when the LAME path is valid, and passes the options field (input and output files are appended; the default is -V2 --noreplaygain, also used if the field is empty). Empty or invalid path falls back to Windows Media Foundation at 128 / 160 / 192 / 224 / 256 / 320 kbps (default 192, CBR). After writing, the app reports whether LAME or Windows ran; failures open a dialog. The written MP3 is not opened in the current tab (only overwriting the MP3 already open keeps that file). Parallel tab exports use the Parallel export count setting (default Auto = a quarter of the CPU cores; the dropdown max is half the cores; Windows MP3 is always one at a time). The last export folder is remembered. LAME is not bundled.
• UI language is also on the gear. The default is Auto (Japanese if the OS is Japanese, otherwise English). The Japanese locale keeps the current wording, including English labels.
• The manual is ?, or GitHub Pages in the browser. It follows the UI language.
• Settings file: %LocalAppData%\MGA\MGA Sonic Anvil\settings.json
• Closing the app (Ctrl+Q, or the window close button) remembers every open tab (including which one was in front), the window position and size (including maximized), and unsaved edits (kept as a working copy), and restores them on the next launch (multiple tabs show the same status-bar progress as opening). Edit history is restored too when the operations can be replayed (Undo back to the original still works). Time / amplitude zoom, waveform selection, and the playhead reset to their defaults. Closing a tab still asks to save. Two or more dirty tabs also offer “Save all and quit” and “Quit without saving any”.
• Only one instance runs. A second start brings the existing window forward and opens the Wave / AIFF / MP3 files you passed as extra tabs.
• On startup the app checks GitHub Releases and tells you if a newer build exists. Yes opens the release page (nothing is downloaded). No skips that version until a newer one appears. Offline, it continues quietly.

12. Scope

• This does not cover everything Interactive Music can do. EXPORT is Wave-only (markers / sample loop).
• In-app preview is an approximation. Dual playheads for Play -E may not match Wwise exactly. Always verify after EXPORT.
• AIFF export is not supported. Choose Wave or MP3. Wave writes 8 / 16 / 24-bit (4-bit is for quantized audition and is saved as 16-bit).
• EXPORT changes the Wwise project. Back up or use version control first.

13. Trademarks and license

• Wwise® / Audiokinetic® are trademarks of Audiokinetic Inc. This tool is unofficial. WAAPI requires a valid Wwise license.
• This app is released under the MIT License.
• LAME is the name of the LAME project and is licensed under the LGPL. This app does not bundle, modify, link, or redistribute it; it only runs a user-supplied lame.exe as a process. Obtaining LAME and complying with its license is the user’s responsibility. This tool is unofficial.
• Signalsmith Stretch is Signalsmith Audio’s pitch/time library (MIT). It is bundled for pitch shift.