Project deliverables
You receive more than a transcript. Each project delivers a structured, traceable set of outputs designed for review, research, archival use and further editorial work.
R2 Mechanics can continue from the source-faithful transcript into speaker and language structure, source reconstruction, restoration, review mapping, chapters, entities, summaries, translation, human quality control and interactive source-linked delivery.
These capabilities can be combined into one connected processing and enrichment workflow according to the needs of the collection.
Layer 1 · Source-faithful baseline
The source-faithful layer preserves what can be established from the original recording while keeping timing, speaker structure, language structure and uncertainty linked to the source. It forms the reference core from which all further processing and enrichment remains traceable. Where required, a separate structured reading version makes the material easier to work with — the two remain clearly distinguished, so editorial refinement never replaces the documented source record.
Source-faithful baseline
This is the documented transcription baseline: speech, speaker structure, timestamps and uncertainty preserved as established during processing. It provides the traceable reference from which all further project outputs are derived.
Editorial version (where applicable)
Where required, a separate reading version can be prepared for efficient research, editorial work or publication workflows. It improves structure and readability while remaining linked by version reference to the source-faithful record.
Editorial refinement supplements the documented source rather than replacing it.
Uncertain passages remain visible instead of being silently normalised. The record shows both what is established and where closer review can add value.
Layer 2 · Analysis & Enrichment
Transcription establishes the first machine-readable layer. R2 can continue from there to reconstruct structure, add analytical and editorial layers and turn audiovisual material into a usable research, archival or institutional record.
The exact combination of layers is defined by the source material, project objective and agreed scope. Not every project requires every available layer.
Difficult material does not have to stop at a low-confidence first transcript.
R2 can add diagnostic audio analysis, restoration or enhancement and comparative processing paths before the transcript and enrichment layers are assembled.
Independent recognition and speaker-analysis paths can examine the same source. Agreements strengthen the record; differences and uncertainty remain visible for review rather than being silently resolved.
R2 can preserve the original-language transcript while adding a separate translation layer in another working language. The translation remains distinguishable from the original transcript and retains the same speaker structure, source timing and navigation context.
This allows multilingual recordings to remain source-faithful while becoming easier to explore across language boundaries.
The translation layer is designed as an aid to research, access and collection exploration. It does not replace the original-language transcript or, where formally required, professional or certified translation.
Where a project requires additional verification or editorial refinement, human quality control can be added to the automated processing and review workflow.
This allows R2 to move from a reproducible automated source-faithful baseline to a thoroughly reviewed research-ready result and, where editorial verification is part of the project scope, toward publication-ready delivery.
Multilingual recordings can remain one continuous source while language regions, speaker structure, transcript content and optional translation are made independently navigable.
Layer 3 · Access & Delivery
The result is delivered as a usable information environment rather than only as a static transcript file.
The structured record is the product. File formats are its delivery forms. Outputs can be combined according to the way your organisation needs to review, preserve or reuse the material.
Interactive HTML
A searchable, self-contained transcript with chapter navigation, speaker structure and direct access to relevant passages. Designed for efficient review of long recordings without working through a continuous document from beginning to end.
DOCX
A professionally structured document with speakers, timestamps and chapters for research, editing, internal review and publication workflows.
Markdown
A lightweight structured format for documentation, repositories, version-controlled workflows and long-term text access.
JSON
Machine-readable transcript data containing timestamps, speaker information and structured metadata for corpus work, integration and further processing.
WARC (where required)
Standards-based archival packaging can be provided where the collection or institutional workflow requires it.
SRT / VTT
Time-aligned subtitle formats for documentary production, media review, accessibility workflows and synchronized playback.
Processing record
A documented project record describing the material processed, relevant processing conditions, selected versions and identified uncertainties. It provides the context needed to understand and review the delivered transcript.
Analysis & Review Map (where applicable)
A visual overview of the recording highlighting regions that require closer attention — for example difficult signal conditions, uncertain recognition or sections selected for additional review. This helps focus restoration and quality-control effort on the parts of the source where it has the greatest value.
Output formats are defined during project scoping. R2 selects the formats that add practical value to the collection rather than producing unnecessary files by default.
R2 can take audiovisual material through source preparation or restoration, transcription, speaker and language reconstruction, analysis, enrichment, translation, human review and interactive delivery as one connected workflow.
The project scope defines which layers are required; the relationship to the original source remains preserved throughout the workflow.
We will help define the outputs that make the material genuinely useful for your archive, research or production workflow.