## Instructions - Free DAW path: LMMS (beats/MIDI sketching, built-in synths ZynAddSubFX) → Ardour (record/mix/master, VST3/LV2, pro routing); both run Linux/Win/Mac, projects in git-ignored `sessions/` + stems export. - MIDI vs audio: MIDI for notes/editing (piano roll, quantize 1/16, velocity 70-110 humanized ±10), bounce to audio (freeze) before heavy FX; audio tracks for vocals/guitar, 48kHz/24-bit session rate. - Arrangement: 8-bar loops → full form (intro 8 / verse 16 / chorus 16 / break 8 / chorus/outro); reference track A/B at matched loudness; cut lows on non-bass (`high-pass 30Hz` master safety). - EQ basics: high-pass vocals 80Hz/guitars 100Hz, cut mud 250-400Hz (-3dB wide), presence 2-5kHz (+2dB narrow) on lead, air 10kHz shelf (+1.5dB); solo in context, not isolation. - Compression: vocals 3:1, attack 10ms/release 80ms, -4dB GR; drums bus 4:1 smash parallel (dry/wet 30%); sidechain bass to kick (duck -3dB, release to tempo) for EDM/pop pump. - Levels/pan: kick -6dBFS peak anchor, bass -8, lead -10, hats -16; pan LCR-lite (drums center, doubles ±40, pads ±60); mono-check lows <120Hz (utility mono). - Reverb/delay: one room (short, drums) + one hall (long, vox sends only), pre-delay 20-40ms, low-cut reverb input 300Hz; delay dotted-1/8 tempo-synced, 15% feedback, low-pass 4kHz. - Master: mix bus -6dB headroom → EQ tidy → glue comp 2:1 (-1dB GR) → limiter ceiling -1.0dBTP → target -14 LUFS integrated (streaming) / -9 LUFS club; dither only on 16-bit render. - Royalty-free sourcing: record/commission first; else Pixabay Music, Uppbeat free tier, YouTube Audio Library, Free Music Archive (check CC-BY), Splice/Cymatics royalty-free packs — save license PDFs per track. - Export: stems 48/24 WAV + MP3 320 for collab, final WAV 48/24 + MP3 + FLAC; filename `artist_title_mix_v3_48-24.wav`; loudness verify with `ffmpeg ebur128` or YouLean free meter. - Room: headphones (ATH-M50x/HD280) + Sonarworks-style correction or cheap treated corner (duvet + bookshelf), mix at 75dB conversation level; rest ears every 45 min. - Versioning: `song_v01_midi`, `_v02_arr`, `_v03_mix`, `_v04_master` saves; notes file per version (what changed); backup sessions + samples to external drive weekly. ## Examples - Lo-fi beat in LMMS: 84 BPM MIDI drums (swing 55%), Rhodes VST chords, vinyl crackle audio, sidechain pad, -14 LUFS master for YouTube. - Podcast-theme in Ardour: 48/24 session, MIDI bass + audio guitar, vocal chain (HPF/EQ/comp/de-ess), limiter to -16 LUFS stereo for spoken-word loudness. - Royalty-safe release: all-melody original + Pixabay shaker loop (license saved), stems + instrumental + clean versions, distributor upload with ISRC. ## Workflows 1. Sketch 8-bar MIDI loop (drums/bass/chords/lead) at target BPM, humanize velocity. 2. Arrange to full form, record audio takes (3 comps → best), comp + tune lightly. 3. Mix: gain-stage, HPF/EQ cuts, compression, pan, send-verbs/delays, automation rides. 4. Reference A/B at matched level, mono + phone-speaker checks, rest + revisit. 5. Master to -14 LUFS / -1dBTP, verify `ebur128`, render WAV/MP3/FLAC + stems. 6. License-log every sample, tag metadata (ISRC/artwork), backup session + stems. ## Templates - Ardour/LMMS session checklist (paste into notes): ```txt SESSION song_v03_mix — 48kHz/24-bit, BPM 92, key Am TRACKS: kick(-6) snare(-9) hats(-16) bass MIDI(-8) rhodes(-11) lead vox(-10) BUS: drums->comp 4:1 | music->glue 2:1 | vox->deess+comp 3:1 SENDS: room 12% (drums) | hall 18% (vox) | dotted-8 delay 10% (lead) REF: __link__ matched -14 LUFS, mono OK, phone OK TODO: ride chorus vox +1dB, tame 300Hz mud, re-check outro tail ``` - ffmpeg loudness verify + convert: ```bash # measure integrated loudness (target -14 LUFS, true peak -1.0) ffmpeg -i mix_master.wav -filter:a ebur128=peak=true -f null - 2>&1 | tail -15 # render distribution copies from 48/24 master ffmpeg -i mix_master.wav -ar 48000 -sample_fmt s16 -dither_method triangular song_16bit.wav ffmpeg -i mix_master.wav -codec:a libmp3lame -b:a 320k song_320.mp3 ffmpeg -i mix_master.wav -codec:a flac song.flac ``` - LMMS MIDI humanize (piano-roll habits): ```txt QUANTIZE: notes to 1/16, swing 50-57% (genre), strength 80% VELOCITY: hat accents 95/70 alternating, snare 100±8, ghost 55 LENGTH: bass legato 95% gap-free, chords staccato 80% for groove CC: mod-wheel swell on pads (0->90 over 4 bars), pitch-bend ±2st max ``` ## Tools - SARAH shell_run for `ffmpeg ebur128` checks, stem converts, file tagging (`id3v2`, `metaflac`); SARAH file_write for session notes + license logs. - LMMS + Ardour + Vital/C Surge free synths + TDR Nova/Kotelnikov free EQ/comp; YouLean Loudness Meter free for LUFS. ## Advanced - Vocal chain order: gain → HPF 80 → de-ess 6k → comp 3:1 (-4GR) → EQ presence → saturate subtle → send hall; double + harmonies ±6 cents detune, ±15ms for width. - Low-end surgery: kick fundamental (50-60Hz) vs bass (80-100Hz) split, sidechain multiband only <120Hz, saturate bass harmonics so phones reproduce root. - Release ops: DistroKid/TuneCore upload WAV + 3000px art, Spotify `-14 LUFS` normalize-on, leave -1dBTP headroom for transcodes; Content-ID only on 100% owned. ## Documentation - https://lmms.io/documentation/ - https://manual.ardour.org/ - https://www.youtube.com/audiolibrary/music - https://pixabay.com/music/ - https://manual.audacityteam.org/ (companion cleanup)