Desktop application · Research workspace

Local transcription and source-linked research

A local-first evidence workflow for moving recorded speech from source to reviewed record, cited extract, and durable artifact.

Relationship
Studio development
Stage
In development
Capabilities
  • Product strategy and management
  • Applied AI
  • Software and data engineering
  • Research and content systems

Availability

An Apple Silicon-focused development build exists. It is not a public cloud service, and generative synthesis remains a gated candidate.

The problem

Recorded speech is easy to separate from its source once transcripts are corrected, excerpts are selected, and derivative materials multiply.

The contribution

JodyInk developed a local-first model that preserves the path from immutable source through reviewed record and anchored evidence to exported work.

Decisions that shaped the work

  1. Make corrections create revisions rather than silently rewriting the underlying source.

  2. Keep transcription local and treat generative synthesis as a separate capability with its own release gate.

What was implemented

  1. Local audio and text intake, on-device transcription, source-linked verification, and offline cited extracts.

  2. Revision-aware evidence, integrity records, durable exports, and an implemented on-device synthesis candidate.