S19 · Pipeline skill
Turn long audio into notes
Organize long recordings and timestamps, review transcripts and produce traceable notes.
Chunked audio transcription + time alignment + human review and summary
- Pipeline
- 5 steps
- Human review
- 2
- Components
- 6
01Pipeline
Base pipeline
Ordered steps. Each consumes the previous output; review gates hold until a person accepts.
- Starts from
- Authorized recording, media metadata and note requirements
- Delivers
- Confirmed notes and recording references
Review contract
People decide at review gates; components supply data, drafts or review surfaces. Unaccepted outputs do not advance.
Integration note
Cross-copy alignment requires supported audio fingerprints; humans verify speaker stitching and action-item assignment.
- 01
Prepare recording sources
- Input
- Authorized recording, media metadata and note requirements
- Output
- Recording list with sources and retention rules
- Podcast and video catalog ingestionMedia metadata normalization + visibility and retention rules
- 02
Transcribe and align
- Input
- Recording list with sources and retention rules
- Output
- Timestamped transcript and copy offsets
- Long-audio transcription, overlap alignment and speaker stitchingLong audio chunk transcription + seam alignment and timestamps
- Ad-aware podcast transcript alignmentAudio fingerprint matching + cross-copy time offsets
- 03
Check text and attribution
Human review- Input
- Timestamped transcript and copy offsets
- Output
- Human-corrected text and speaker references
Decided byAssigned reviewer or requesting user
- Transcript translation and viewpoint extractionTimed text translation + speaker and source alignment
- 04
Draft the notes
- Input
- Human-corrected text and speaker references
- Output
- Source-linked note draft
- Grounded document summarizationSource text summarization + citations and fallback labels
- 05
Review and export
Human review- Input
- Source-linked note draft
- Output
- Confirmed notes and recording references
Decided byAssigned reviewer or requesting user
- Office/PDF text extraction, scoped storage and document authoringBounded document extraction and authoring + previews and structural checks
02Industry versions
Industry versions
Healthcare Transferable use Turn consented appointment-service interviews into timestamped notes for administrators to verify names and dates.
Adaptation Service records only, not diagnosis or automated clinical decisions; restrict retention and access.
- 01
Register interview sources and consent
- Input
- Consented appointment interview recordings and administrative note requirements
- Output
- Interview recordings within consent scope
- Podcast and video catalog ingestionMedia metadata normalization + visibility and retention rules
- 02
Transcribe chunks and locate passages
- Input
- Interview recordings within consent scope
- Output
- Timestamped interview text
- Long-audio transcription, overlap alignment and speaker stitchingLong audio chunk transcription + seam alignment and timestamps
- Ad-aware podcast transcript alignmentAudio fingerprint matching + cross-copy time offsets
- 03
Administrator checks attribution
Human review- Input
- Timestamped interview text
- Output
- Checked names, dates and statements
Decided byAssigned reviewer or requesting user
- Transcript translation and viewpoint extractionTimed text translation + speaker and source alignment
- 04
Draft administrative notes
- Input
- Checked names, dates and statements
- Output
- Service notes linked to recording passages
- Grounded document summarizationSource text summarization + citations and fallback labels
- 05
Confirm and archive notes
Human review- Input
- Service notes linked to recording passages
- Output
- Confirmed administrative notes and replay references
Decided byAssigned reviewer or requesting user
- Office/PDF text extraction, scoped storage and document authoringBounded document extraction and authoring + previews and structural checks
03Parameters
Parameters
From the pack’s params.schema.json. Choose values to build a synthetic request.
skills/_pipelines/long-audio-to-notes/params.schema.json
{
"synthetic": true,
"run_id": "synthetic-example",
"locale": "en",
"industry": "base",
"adapter_mode": "local",
"records": [
{
"id": "synthetic-record-1",
"synthetic": true,
"text": "Synthetic observation A: a draft needs review."
},
{
"id": "synthetic-record-2",
"synthetic": true,
"text": "Synthetic observation B: retain the unresolved question."
}
],
"decisions": {}
} 04Member components
Member components
-
Podcast and video catalog ingestion
Media metadata normalization + visibility and retention rules
Working 1 pipeline -
Long-audio transcription, overlap alignment and speaker stitching
Long audio chunk transcription + seam alignment and timestamps
Working 1 pipeline -
Ad-aware podcast transcript alignment
Audio fingerprint matching + cross-copy time offsets
Working 1 pipeline -
Transcript translation and viewpoint extraction
Timed text translation + speaker and source alignment
Working 1 pipeline -
Grounded document summarization
Source text summarization + citations and fallback labels
Working 1 pipeline -
Office/PDF text extraction, scoped storage and document authoring
Bounded document extraction and authoring + previews and structural checks
Working 2 pipelines
05Skills pack
Skills pack
Runnable reference on synthetic fixtures with tests. Domain contracts, services and approvals still need integration.
long-audio-to-notes-1.0.0.zip
Contains SKILL.md, pipeline.json, params.schema.json, runner, tests, fixtures, integration checklist and sources.
Synthetic workflow reference; service integration pending
Download · 18KBSHA-256 a6bb4737d2c18c962c60d1071c025eb3b926bb3407be42bc4dd544bc257b5909