The same rent-vs-own split that hit sample libraries is hitting AI music tools right now. The producers who own their sample catalogs kept shipping. The producers who rented from Splice kept paying. The same fork is open in 2026.
Every indie release is a sequence of five jobs. The 2026 toolchain makes each one cheap enough that one person can run the whole pipeline from prompt to Bandcamp.
Suno and Udio can produce a full song from a 50-word prompt in 60 seconds. They're the new demo tool. You use them to lock the vibe before you commit to a 4-hour production session.
magenta-rt is 2.4B params, on-device via MLX on Apple Silicon, open-weights. PyPI package `magenta-rt` (latest 2.0.3, 2026). The model is yours, the output is yours, the fork is yours. MusicVAE is the older Python path — the live line is the JS-side `@magenta/music` if you want browser-based trio / melody continuations.
Ableton for the arrangement and the live set. EWI for the breath, the bend, the human layer that AI still can't fake. The combination is what makes the track sound played, not generated.
Vital is a free wavetable synth that competes with Serum. Spitfire LABS is a free sample library that competes with anything below $200. The "indie producer tax" is over.
DistroKid for the streaming distribution. Bandcamp for the indie-friendly revenue share (~82% to artist, no DSP gatekeepers). LANDR for the AI mastering that ships the track loud without crushing it.
Splice raised prices three times in five years. The producers who owned their sample catalogs kept shipping. Suno is the next Splice. The fork is: do you own the model, or do you rent the inference?
Suno's prompt is a structured list of genre, mood, instrumentation, tempo, and vocal style. The more specific you are, the more Suno surprises you. The more vague, the more it sounds like Suno. The template that works.
# ~50 words. 4 fields. Vocals at the end. [GENRE] dark ambient trap, lo-fi soul, whispered hook [MOOD] hollow, resigned, slow burn [INSTRUMENTS] sub-bass 808, vinyl crackle, sparse piano, breath texture [TEMPO / KEY] 68 BPM, F minor [STRUCTURE] 8-bar intro (piano + crackle) / 16-bar verse (rap, half-time) / 8-bar hook (whispered, layered x3) / 8-bar bridge (strip to bass + voice) / 8-bar outro (reverse piano tail) [VOCAL STYLE] close-mic rap verses, airy layered whisper hook [REFERENCES] Massive Attack restraint × Earl Sweatshirt cadence [LYRICS HOOK] "I left the room but the room left me"
The worked example: a real session, real timestamps, real tool sequence. Five days, four tracks, one Bandcamp release. The same five layers, run end-to-end.
EP: "Closed Loops" — 4 tracks, ~22 minutes total, lo-fi + dark ambient + whisper-vocal. Audience: Bandcamp + SoundCloud. Budget: $0 in tools (everything indie-tier), $0 in marketing. Goal: ship by Friday.
Mon: Suno vibe demos (4 tracks, 2 hours). Tue: magenta-rt MIDI foundations. Wed: Ableton arrangement + recording. Thu: Vital sound design + EWI tracking. Fri: LANDR master + Bandcamp release.
# Real session, by day # Monday · 2 hours · Suno for vibe demos (4 tracks) Suno v4 prompt → 8 variations per track → 4 winners # ~$0 in credits Export stems (vocals, drums, bass, other) for each winner # Suno Pro tier Drop stems into Ableton for arrangement reference # mute originals, build over # Tuesday · 6 hours · magenta-rt for the foundation MLX launch → load magenta-rt checkpoint (2.4B params) # ~10s cold start on M3 Pro Prompt: 16-bar chord progression in F minor, 68 BPM # MIDI out, copy to Ableton @magenta/music → generate 4 melody variations per track # pick the one that "feels right" # Wednesday · 8 hours · Ableton arrangement + tracking Ableton Live 12 → arrange MIDI + record EWI solo # EWI 5000 + breath controller Spitfire LABS strings on the choruses # free, sounds like a string section Ableton comping → pick the best of 4 takes per section # takes are 2-3 min apart # Thursday · 6 hours · Sound design + EWI + vocals Vital → design 2 patches: sub-bass + airy pad # 90 min, both patches saved EWI → 3 improvised solo passes per track # the human layer Whisper-vocal hook → 5 layered passes, B3 compressor # close-mic SM7B, no autotune # Friday · 3 hours · Master + release LANDR mastering → "Loud" preset, -9 LUFS # $4/mo, ready in 8 minutes Bandcamp → album art + tracklist + $7 price tag # ~82% to artist, no DSP gatekeeper DistroKid → schedule streaming for next Friday # $22.99/yr, all platforms Total: 25 hours across 5 days. Tool spend: $4 (LANDR) + $0 (everything else). Shipped: 4 tracks, 22:14 runtime, 38 pre-orders in week one, 14 on Friday launch.
The five jobs are universal. The 12 tools are 2026. Swap tools as the field moves; the layer model holds.
Nine stacks. One philosophy. Zero babysitting.