How we work

What happens to your material — step by step.

A controlled process from intake to delivery — documented at every stage, with no silent automation.

Processing difficult archival audio is not a single operation. It is a sequence of controlled steps, each building on the last and remaining reviewable throughout.

This is what happens to your material when you work with R2 Mechanics.

Step 1

Scope & Protect

Before processing begins, we define what material we receive, how it is handled, what will be delivered and under what confidentiality conditions.

Every project starts with a written scope. We do not begin without one.

Step 2

Preserve the Source

A controlled processing copy is created from your source material. The original is never overwritten or modified and is not used as the working processing file.

Every copy created is documented.

Step 3

Analyse with Independent Paths

Your material is analysed through multiple independent recognition and speaker-analysis paths. Each contributes evidence. None is treated as automatically correct.

Where independent paths agree, the evidence becomes stronger. Where they differ, the difference is documented — not silently resolved.

Speaker attribution is handled as a dedicated analysis step. It is not a byproduct of transcription.

Step 4

Review and Select

Evidence is compared. A documented working baseline is selected through a controlled, traceable process.

No automatic finalisation. No silent selection. The chosen transcript is documented, reproducible and operator-reviewed.

Step 5

Structure and Deliver

The validated transcript is transformed into structured, navigable outputs. If editorial refinement is required, it is a separate, documented stage — it does not alter the forensic baseline.

You receive the source-faithful record alongside any production-adapted versions, with the two kept clearly distinct.

Our principle

Evidence before polish.

The forensic foundation and any editorial layer remain separate. You always know what formed the evidential basis and what was derived from it.

When languages shift.

Recordings where languages, speakers or recording conditions shift require adapted processing strategies.

For multilingual and mixed-language material, we assess the specific conditions of your collection before processing begins. We adapt the analysis to your material — not the other way around.

Questions about how your material would be handled?

Discuss your collection →