Project deliverables
Every project delivers a documented, structured set of outputs. Here is what you receive — and what each output is designed for.
Every project produces a forensic baseline and, where needed, a structured reading version. They are never merged — what was found and what was refined stay separate.
Forensic baseline
The unmodified result of the analysis — what the processing found, preserved as-is. This forms the primary source-faithful project record. It is the foundation from which everything else is derived, and it is never overwritten.
Editorial version (where applicable)
A separately produced version adapted for reading, review or publication — where project scope requires it. Clearly marked as derived from the forensic baseline, not a replacement for it.
Interactive HTML
Chapter navigation, speaker filtering and full-text search in a self-contained file. Designed for archival review, research and institutional handover.
DOCX
Formatted document with speaker labels, timestamps and chapter structure. Suitable for standard editorial, review and documentation workflows.
Markdown
Plain-text structured format for version-controlled archival, documentation pipelines and research repositories.
JSON
Structured data with per-segment timestamps, speaker IDs and metadata. Suitable for downstream processing, corpus work and data integration.
WARC (where applicable)
Standards-based archival packaging for project files and metadata where required by the collection or institutional workflow.
SRT / VTT
Subtitle-format exports aligned to source timing. For media synchronisation, review, accessibility workflows and documentary production.
Processing record
A documented summary of how your material was processed — inputs, conditions, versions selected and any uncertainty flags. Included as a standard project deliverable.
Format availability is confirmed during project scoping. Not every format is appropriate for every collection type.
Written project scope before processing begins
What material is received, how it will be handled, what will be delivered and under what conditions. Agreed before we start.
Speaker-attributed, time-stamped transcript
Speaker attribution is handled as a dedicated analysis step — not as an incidental output of transcription.
Chapter-level structure
Long recordings are segmented for navigation. Chapter boundaries are derived from the structure and content of the source material rather than imposed arbitrarily.
Documented uncertainty
Passages where recognition confidence is low are marked — not silently corrected or omitted. You know where the limits are.
Processing record
A documented account of what was processed and how. Delivered with the transcript so the output can be understood, reviewed and, where necessary, independently assessed.