Tools you can actually use
Every cluster ships something you can actually open: live, beta, or an honest prototype. We don’t hide products behind a "Join the waitlist" button.
Seven labs · One platform · Building in public
The overlap between audio and AI deserves a real home.
It starts with analysis you can trust. Transparent DSP, verified against the EBU R128 reference, running in your browser so your audio never leaves it. We’re building the AI layer on top: one that learns to read audio and turns the numbers into guidance.
BS.1770-4 LUFS · WebAudio API · AAudio · Realtime DSP · Web Speech
51 minutes of audio analyzed · 19 tracks
01 Labs
Seven labs.
One coherent platform.
Each lab owns a specific corner of audio and AI, from mix feedback to hearing support to live routing. Same brand, same rigor, real products.
Free in your browser · paid at scale
The same engine, as an API.
Every lab runs free, on your device. Need that same analysis server-side, in batch, or inside an AI agent? The hosted API and MCP server put loudness, voice, and signal analysis behind one key, verified against EBU R128 / BS.1770-4.
A working studio, not a landing page
Audio is craft. AI is leverage.
The platform should respect both.
02 What we don’t do
People-first writing
No keyword-stuffed AI slop. Every page is written by someone who has shipped audio in the wild and is willing to be wrong on the record.
Built where it matters
Android, realtime DSP, accessibility, and developer ergonomics. The places audio AI actually has to work, not just generate marketing demos.
03 Featured tool
MixLab · Live demo
MixLab Analyzer.
MixLab Analyzer.
Real WebAudio, in the browser.
Drop in a track. Get readable numbers (dynamics, brightness, stereo width, low-end density, harshness risk) and short feedback cards in plain language. No upload to a server, no model in a box, no "magic enhance" button.
- FFT-based spectrum analysis with peak and RMS detection
- Mid/Side decomposition to estimate true stereo width
- Crest-factor and bandwise dynamics scoring
- Runs entirely client-side. Your audio never leaves the page
mixlab/analyzer · BS.1770-4
reference_master.wav Waveform · 3:24 48 kHz · stereo
Short-term LUFS · 3 s window integrated: −14.2 LUFS
Integrated LUFS
−14.2 BS.1770-4
True peak
−0.8 dBTP
Loudness range
5.8 LU
Crest factor
8.6 dB
Stereo width
0.71
Centroid
3.24 kHz
Spectrum · log frequency · normalised
A
Tonal balance
Sub 20–60 62%
Bass 60–200 92%
Lo-mid 200–500 78%
Mid 500–2k 85%
Hi-mid 2–4k 74%
Presence 4–8k 66%
Air 8–16k 42%
Harshness 18%
Plain-language read
Sits in the streaming pocket. Dynamics are modern but not flattened. Bass shelf is healthy; watch the 2–4 kHz band on the next pass, presence is starting to bite. Mono-safe with stereo width to spare.
04 Latest insights
Deep dives, not blog posts.
If you build, ship, or rely on audio, this is for you.
Tools, docs, demos, and product lines covering the parts of audio AI that the big platforms don’t bother with. No newsletter capture before you see the goods.