nodedsp:~$ ./spectral –launch
live now · spectral.nodedsp.com

spectral.

Do everything to a single audio file — sound design, sampling, restoration and mastering, in one browser tab.

A scientific-grade audio workstation with an assistant you drive in plain english. Nothing to install.

runs in your browser · free editor, no account needed to try

$ spectral –what

One file. Every possible thing you’d want to do to it.

Drop in any audio and Spectral opens a full ITU / EBU / ISO-compliant analysis report — loudness, dynamics, stereo field, frequency, THD+N, key, transients, clipping. From there you can edit it by hand in a browser DAW, reshape time and pitch like a VariPhrase, slice it into a sampler instrument, surgically erase sounds off the spectrogram, split it into stems, repair it, and master it to spec. The editor, engine, loops and metering are free forever — no length caps, no format locks. Subscriptions pay for the cloud: AI, GPU jobs and sync.

$ ls capabilities/

The full toolset

~36 processing tools across the whole signal chain. Everything below is real and shipping today.

analysis

Scientific report on every file

  • ITU-R BS.1770-4 / EBU R128 loudness
  • dynamics, crest & dynamic range
  • stereo correlation, width, M-S
  • 1/3-octave (ISO 266), THD+N, SNR
  • key detection & tuning, transients, clipping
edit

Sample-accurate editor

  • hi-res waveform + spectrogram
  • trim, split, silence, fades, invert
  • cut / copy / paste regions
  • full undo / redo history
  • in-browser recording & live input
time & pitch

VariPhrase-style manipulation

  • Time Trip — scrub through frozen time
  • warp markers, elastic re-timing
  • beat grid, BPM, snap & bar loops
  • pitch & formant shift ±24 st
  • Paulstretch extreme stretch
sampler

Slice to instrument

  • slice by bars, beats or transients
  • audition on pads, up to 128 slices
  • export kit — 24-bit WAV + SFZ
  • chromatic keymap from C1
real-time fx

Master Section chain

  • EQ, compressor, delay, reverb
  • chorus, flanger, phaser, width
  • LFO modulation + automation record
  • reorder, bypass, then bounce
spectral

Photosonic eraser

  • draw on the spectrogram to remove
  • lift a cough, a beep, a bleed
  • adjustable cut depth, fully undoable
master

Metering & loudness

  • live M / S / I / LRA / TP meters
  • normalize to peak, RMS or LUFS
  • EBU R128 export normalize
  • −14 LUFS default target
$ spectral –workspace

Analysis on the left. Your waveform in the middle. The assistant on the right.

One screen holds the whole job — the scientific read-out, the editor and its meters, and the plain-english assistant, side by side.

spectral — workspace● −14.0 LUFS
analysis
loudness−14.0 LUFS
true peak−1.0 dBTP
dyn range11.4 dB
stereo+0.72
keyA minor
tempo124 bpm
thd+n0.004%
clippingnone
phaseok
waveform
trimslicetime triperasefx chainstems
M
S
I
LRA
TP
assistant
you
clean up the hum and get it to −14 LUFS
spectral
on it — building a pipeline:
1 de-hum · 60 Hz + harmonics
2 denoise · light
3 normalize · −14 LUFS
run pipeline

a representation of the live workspace — the real thing is at spectral.nodedsp.com

// signature

Talk to it. Watch it build the pipeline.

Describe what you want — “clean up the hum, tighten the low end, get it to −14 LUFS” — and the assistant assembles a multi-step processing pipeline, previews each move, and executes it with full undo.

Bring your own model — Claude, GPT-4o, or a fully local Ollama. NodeCloud & up.

// signature

Split any mix into stems.

Demucs separation runs on NodeDSP compute nodes and hands back vocals, drums, bass, guitar, piano and more — each one solo-able, re-mixable, and importable as a fresh editable file.

Reopen a track and the stems are already there. NodeCloud & up.

// signature

PURE — a bit-exact reference path.

One button hard-bypasses the entire effect chain: a float32, untouched-samples signal path for critical listening. The kind of trust a mastering engineer needs before they’ll believe anything else on the screen.

Free, on every session.

$ spectral –pricing

Pay for the cloud + AI. Your device is free.

The whole editor — waveform, FX, loops, VariPhrase time tools, and broadcast-grade metering — is free forever, with no file-length cap and every export format at any sample rate. A subscription buys the things that spend our dollars: the AI assistant, stem separation and restoration on NodeDSP compute nodes, and cloud sync for your projects.

Capability Freethe editor, forever NodeCloud$10 / mo NodeDSP License$25 / mo
Editor, FX chain, loops, VariPhrase, metering
Slice-to-sampler & kit export
File lengthno limitno limitno limit
Export formats & sample ratesallallall
AI assistant1,000 msgs / mo3,000 msgs / mo
Stem separation (GPU)30 / mo300 / mo
Restoration suite
Cloud syncon-device only5 GB50 GB
NodeWave on the weblocal only
Native NodeDSP engine & future instrumentsincluded as they ship
API access (MCP) & priority queue
Full supportcommunitycommunity

Every on-device tool is unpaywalled on purpose — the free tier is a genuinely great editor with no caps. You only pay when a job leaves your machine: our GPUs, an LLM, or cloud-synced storage.

$ spectral –why-trust-it

Built for people who measure things.

◐ bit-exact

PURE reference path

Hard-bypass float32 pass-through for critical listening — no processing in the signal path.

▮ broadcast

BS.1770 metering, free

Real ITU / EBU loudness — M, S, I, LRA and true-peak — live, and never paywalled.

◇ open dsp

MIT-licensed engine

Signalsmith stretch, Demucs, Paulstretch — high-end algorithms, no proprietary black boxes.

⌘ agent-ready

Yours to automate

Bring your own AI key, or drive Spectral from any MCP agent with the NodeDSP License.

follow the signal.

Spectral is live. Open a file and see how far the rabbit hole goes.

nodedsp — audio processing, done in the open
spectral.nodedsp.com