AUDIOSHAKE REFINERY – AUDIO DATA FOR AI TEAMS

Make any audio trainable.

The Refinery turns raw, mixed recordings into clean, training-ready data — diagnosed, separated, scored, never synthetic. Bring us the audio you've licensed, acquired, or created, and we return isolated, confidence-scored stems your models can actually learn from.

Data infrastructure for
Frontier labs
Speech & voice-AI teams
DATA MARKETPLACES
CONTENT OWNERS
80+
languages supported by AudioShake's sound separation infrastructure
500M
minutes of audio processed to date
Trusted by
The world's largest AI labs, frontier model teams, and data marketplaces refine their audio here.
01 – THE REFINERY

One engine. Any audio in, training-ready data out.

Most audio a lab licenses and acquires is a black box — multi-party, uneven, and difficult to trust at scale. The Refinery tells you what's in it, what's good, and then makes it usable.

Step 01 — Diagnose

See inside the black box

Point us at your corpus. We assess every hour for what actually limits training — background music, speaker overlap, noise, degradation, bleed — and determine which audio separation models will deliver the best results.

Step 02 — Separate

Isolate the real signal

Source separation pulls clean stems out of finished, mixed recordings — dialogue, per-speaker tracks, music-free speech — from a single track, no session files needed.

Step 03 — Score

Put a number on quality

Every output ships with confidence scores, so you filter, weight, and QC your dataset programmatically instead of listening your way through it.

Step 04 — Deliver

Ready to train

Structured, training-ready data delivered over API or inside your own environment when the audio can't leave your walls.

02 – REFINE

Your audio, refined.

Refine cleans and scores what you bring. Send us the recordings you've licensed, acquired, or created — even if you don't yet know what's usable. We diagnose the whole corpus, then return clean, isolated, confidence-scored stems, ready to train.

Diagnose what's usable, fixable, or junk before you spend compute on it

Works from finished, mixed recordings — no original stems required

Music, noise, and bleed removed; speakers split into individual streams

It's your real audio, isolated — never synthesized or hallucinated

→ Output: scored, separated, training-ready
For frontier & voice-AI labs

Turn licensed archives into training data

Turn the licensed archives you already own into training-ready data for ASR, diarization, speaker-ID, and TTS. Diagnose quality before you spend compute and labeling budget, and separate everything you need from finished recordings — so your existing corpus becomes a data source.

For data marketplaces

Upgrade raw inventory into sellable datasets

Upgrade raw audio inventory into isolated, scored, training-ready datasets you can sell. Refine adds the separation and quality layer buyers trust — turning mixed recordings into structured, per-source data with confidence scores attached.

03 – AUDIOSHAKE MODELS

Usable, structured data.

Send us finished, mixed recordings — no original stems or multitrack sessions required. We return training-ready components, each isolated from the real signal.

Spoken dialogue pulled out of real-world audio — crowd environments, on-location recordings, mixed broadcast content. AudioShake's various speech isolation models work across a range of resolution and content types.

Film: “Hidden in Plain Sight” — Gregg Dunham & Mason Frenzel
Dialogue Isolation
0:00
SPEAKER OVERLAP
Multi-Speaker Separation
View product page

Multi-speaker conversation – including overlapping speech and filler words – split into individual speaker tracks.

Film: “Meridian” — Netflix Open Source, CC Attribution
Multi-Speaker Separation
0:00
MUSIC
Instrument Stems
View product page

Isolate 10+ individual instruments — vocals, bass, drums, guitar, lead and backing vocals, and more — from fully mixed tracks.

Stem Separation
0:00
MUSIC
Music Removal
View product page

All background music removed — including sung vocals — leaving a pure speaker signal where music would otherwise contaminate the training set.

Film Credits: Jaywalker Music
Commercial Music Removal
0:00
04 – CONFIDENCE SCORES

Know the quality before you train.

Every multi-speaker output ships with confidence scores — signals indicating how cleanly a source is isolated across the track. Filter, weight, or QC your dataset on numbers, not subjective listening.

Watch how confidence scores work →
Speaker 1
0.96
Speaker 2
0.91
Speaker 3
0.62

Drop the 0.62 stem, or route it to human QC — before it reaches your pipeline.

Customer
/
Luel

We collect audio in the real world, which means real-world noise and interference. AudioShake has helped us process thousands of hours of clean, speaker-separated data, enabling the world's leading labs to build better models.

William Namgyal
CEO & Co-founder, Luel
05 – WHY AUDIOSHAKE

Built for the way labs actually train.

01

Diagnose before you spend

We assess your corpus upfront to determine the separation models that will get the most out of it — so compute and labeling budget go to audio that will actually improve training, not audio that was never going to help.

02

From a single mixed track

No original stems, no multitrack sessions, no isolated recordings required. We separate finished audio using our state-of-the-art models — which means your existing archive is already a data source.

03

Never generative

Every stem is real audio, isolated — not synthesized, not hallucinated, not filled in. What you train on is exactly what was recorded.

04

Consistent and repeatable

The same input yields the same output at scale, so no distribution shift creeps in from inconsistent preprocessing — the failure mode that destabilizes benchmarks.

05

Cloud or on-prem

Run on our API, or deploy inside your own environment when the data can't leave your walls.

06

Cleared to train

Your audio never becomes someone else's dataset. We process what you bring and return it to you — and in on-prem deployments, it never leaves your environment. Provenance you control.

06 – FAQs

Frequently Asked Questions

What do we get back?

Structured, training-ready components — isolated stems and per-speaker tracks, each with confidence scores — delivered in our cloud or your environment.

We're a data marketplace — can we run Refine on inventory we resell?

Yes. Marketplaces use Refinery to turn raw audio inventory into isolated, scored, training-ready datasets with a quality layer buyers can trust.

Can your technology deploy inside our own environment?

Yes. The Refinery can be deployed on-prem so audio never leaves your walls.

Who owns the data and the outputs?

You do. The Refinery processes the audio you bring and returns it to you; we don't repurpose your data and all data is deleted upon completion of the job. The Refinery can also be deployed on-prem.

Do you need our original stems or session files?

No. AudioShake's Refinery works from finished, mixed recordings — a single track is enough.

Why does separated audio produce better training data than raw mixed recordings?

Mixed audio entangles signals — overlapping speakers, music, and noise — so a model can't cleanly learn any one of them. Separated, per-source stems give your model the isolated signal it's actually meant to learn.

Can AudioShake process audio at the volume an AI training pipeline needs?

Yes. The Refinery is built for corpus-scale throughput; we've processed over one billion minutes of audio to date.

What if we don't know which of our audio is worth training on?

That's what the Diagnose step is for. We assess your whole corpus and flag what's usable, fixable, or junk before you commit compute or budget.

07 - CONTACT US

Enter the Refinery

Company type
Pick all that apply
I'm interested in *
Pick all that apply
Select at least one option.
Data type
Pick all that apply
* Required
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.