What is AudioLab.tools?

AudioLab.tools is an independent measurement lab for speech AI. We measure speech audio independently, with numbers you can re-run and check, and we build the tools to keep measuring it yourself: audits of your recordings, AudioLab Local on your own machine, VeldLab for measurement days, seven free browser labs, an API and an MCP server. We sell no speech recogniser, noise suppression or text-to-speech.

Is AudioLab.tools an app or a piece of hardware?

Neither. AudioLab.tools is a browser-based platform: open the URL and the tools run in your tab, nothing to install. It is not affiliated with any mobile audio-recording app, audio editor, or hifi-equipment brand that shares the name.

Does my audio get uploaded to a server?

Not in the browser labs. The analysis tools run entirely client-side using the WebAudio API. Your audio is decoded and processed in your own browser and never leaves your machine. Two exceptions: HearLab’s live captions use your browser’s speech recognition, which may send the audio to an online speech service (Chrome does). And the hosted API and MCP server receive the audio you send them, process it on EU servers and discard it within the hour at the latest; the privacy page has the details.

How accurate is the loudness measurement?

MixLab implements the ITU-R BS.1770-4 / EBU R128 loudness algorithm. Against the EBU Tech 3341 test cases its integrated loudness lands within 0.06 LU of the standard’s expected values, inside the ±0.1 LU tolerance the standard allows, and every test case, expected and measured, is published on the API page. The full method, filters and standards are documented on the methodology page.

Who builds AudioLab.tools?

AudioLab.tools is built and maintained by Nathan Renting, a working audio developer based in Eindhoven, the Netherlands. It is solo-built and self-funded, with no investors.

Is AudioLab.tools free to use?

Yes. The browser-side demos are free to use with no account and no waitlist. Some labs plan premium tiers for batch and team features, but the core tools are usable for free today.

The independent measurement lab for speech AI.

We measure your speech audio independently, with numbers you can re-run and check, and we build the tools to keep measuring it yourself. One engine in your browser, your pipeline and on your own machine: 157 measurements, each with a status, and a margin where we measured one; its loudness is verified against the EBU R128 reference.

Independent: we sell no speech recogniser, noise suppression or text-to-speech.

So we have no stake in what the measurement says. AudioLab is built by Nathan Renting in Eindhoven, an audio engineer who works on both sides: the measurement chain and the evaluation of speech recognition. Self-funded, no investors.

What: independent measurements of your speech audio, with numbers you can re-run.

Not only the basics (level, noise, clipping, bandwidth), but the whole path a recording takes to a recogniser, and, as a project on request, how recognition holds up on names, addresses, phone numbers and digit strings. 157 measurements, each with its status on /lenses, and a margin where we measured one.

How: we measure it for you, or you measure it yourself.

An audit of your recordings at a fixed price; AudioLab Local on your own machine; VeldLab for a measurement day with calibrated playback. Or start free in the browser and automate the same checks over the API and MCP.

Start here: free, in your browser

Measure a speech recording →

Work with the lab

An audit of your recordings Fixed price, from €490 excl. VAT. Every finding with its rule and number. AudioLab Local (beta) The measurements on your own machine; your audio is not uploaded. Evaluation after an NDA. VeldLab (beta) A field-measurement lab for a measurement day with calibrated playback. Pilot on request.

Automate the same checks: API and MCP

Voice AI

Building or testing a voice agent or speech recognizer? Measure a recording Stress-test your recogniser Batch or API

Thank you for being early: the first 50 accounts from 19 October get the Pro version of one of our Android apps on us, the first 20 the whole series. Yours to pick on your account page. /account →

speech-female-pangram.mp3 · 15.1 s Example report

Level and clipping

-23.2 dBFS

Speech level

Validated

Noise

36.5 dB

Signal to noise

Validated

Margin: ±6 dB

Gaps in the audio

Dropouts

Validated

Pitch and voice quality

218 Hz

Pitch

Validated

Margin: ±3 percent

See the full report →

Real output of the engine on one sample recording, with the status of every number. Your own recording gets the same report, in your browser.

The loudness is verified against EBU R128 / BS.1770-4; every other measure carries its own status.

Seven free labs

MixLabyour track VoiceLabyour take HearLabhard hearing moments SignalLabwhat is in a recording CueLablive audio SkillLabtrain your ears Ψ Labmorph sound

Android apps

RoomLabyour room MicLabyour mic LevelLabloudness for platforms NoiseLabhow loud is it here SignalLabtest signals TuneLabtuning SkillLabear training

Also show prices in A guide at the ECB rate of 19 October 2026. You are always billed in EUR.

How we know the numbers hold

Checked: loudness within 0.06 LU of the EBU Tech 3341 test cases, inside the 0.1 LU the standard allows.
Repeatable: AudioLab Local gives byte-identical output on two runs over the same files, with SHA-256 hashes in a manifest, so a third party can check the result.
Every build: pricing, API docs, privacy pages and claims are tested against the code; a separate sweep breaks things on purpose to prove each check can fail.

The evidence, area by area →

The same engine as an API and MCP server: one key, the free tier is 1,000 calls a month. /api

13 hours of audio analyzed · 358 tracks

Articles on loudness, microphones and noise

About AudioLab: solutions, tools, roadmap, press

Methodology Solutions Tools Roadmap About Press