--- name: source-transcription description: Establish or proof a source-language transcript against the actual recording. Use for dialogue-list preparation, shared transcript verification, uncertain speech, speaker identification and source assets reused across translation locales. --- # Source Transcription Read the [shared expert contract](../.shared/references/expert-contract.md). Acknowledge this skill, actual actor identity, remit, received assets and gaps. Work from observable recording evidence; a plausible script is a reference. ## When to invoke Use when recorded speech must become a reliable shared source, when a supplied transcript needs verification, or when a linguist raises an uncertain passage. You can work directly from an authorized recording and transcription brief. ## Inputs and evidence - Exact recording/cut and selected soundtrack, language, duration and time basis. - Authorized playback/listening tools with actual access to the commissioned audio. - Scope: verbatim or edited source, speaker information, annotations and delivery. - Scripts, names, terminology and earlier transcripts with their provenance. - Required source-language competence and escalation route for uncertain speech. ## Workflow 1. Verify the asset identity and that the selected audio can actually be heard through available tools. Note language switches, channel problems, difficult sections and mismatches between claimed and observed duration or content. 2. Establish the commissioned coverage. Use stable passage/event identifiers and time references anchored to the recording where the tools support them. Preserve original references so later corrections can be located reliably. 3. Transcribe or compare the supplied transcript with the recording in context. Revisit difficult speech, speaker changes, names, numbers, negation and code switching. Check scripts for improvisations, omissions and dialogue absent from this cut rather than assuming the script matches the performance. 4. Research names and ordinary terminology against approved references. Mark unresolved speech as uncertain with location and plausible alternatives. Seek another source-language listener, clearer audio or production guidance through the coordinator when available evidence cannot settle the passage. 5. Proof the complete assigned transcript against the selected recording. Keep distinctions between what was heard, what was inferred and what the client supplied. An automated transcript is a draft until that comparison is done. 6. Deliver the verified source and uncertainty record to the coordinator and downstream linguists. If a shared source changes, identify affected templates and locales; route a replacement immutable translation job through the [runtime guide](../.shared/references/runtime.md). If only a shooting script or extracted text is available, return a source verification gap and a proposed listening task. Do not label the text an observed transcript or invent recording timestamps. ## Outputs and handoff Use the [expert handoff](../.shared/templates/expert-handoff.md). Supply transcript or dialogue list, asset/audio identity, assigned and actually checked coverage, speaker convention, uncertainty locations, research sources and open questions. Separate newly authored text from an independent source-proofing pass when both are commissioned. Tell spotting, translation and terminology specialists exactly which source revision is verified and which passages remain provisional. ## Completion evidence The assigned source coverage was actually listened to and compared. Every uncertain passage is either resolved with evidence or retained explicitly with its impact. A script-only or automated-only pass remains labeled as such. ## Must not Do not guess unintelligible dialogue, invent speaker identities, claim audio verification from text alone or treat repeated inference as corroboration. The bundled text MCP runtime does not transcribe or play recordings; obtain suitable authorized tools before claiming this work ran.