Intelligence · Component · long-audio-asr-assembly
Long-audio transcription, overlap alignment and speaker stitching
What it does
Chunk long audio, transcribe bounded jobs, reconcile seams and assemble timed segments with usage records. One variant uses queue/byte-range cuts; another uses FFmpeg/parallel requests. Speaker stitching is heuristic and varies by source.
Core pattern
Long audio chunk transcription + seam alignment and timestamps
- Input
- Publisher audio URL and episode metadata
- Output
- Transcript, speaker map, segments and spend/job records
Key parameters
| Name | Type | Default | Values |
|---|---|---|---|
| audio_url | string | — | — |
| language | string? | — | — |
| chunk_seconds | number | 460 | 10–480 |
| overlap_seconds | number | 20 | 0–60 |
| stitch_speakers | boolean | true | — |
Maturity and limits
Source paths were inspected statically; tests and live integrations were not executed. This rating describes implemented code, not production validation.
No root reuse license found; third-party news, research, logos and personal profile data require appropriate rights; external services require credentials. Speaker bridging is heuristic; publisher Range behavior varies. Overlap alignment and speaker mapping are heuristic rather than biometric speaker verification; byte-range extraction requires compatible MP3 or unencrypted AAC-in-MP4 layout. Variant-specific dependencies, limits and weaker maturity are recorded under variants; these implementations are alternatives, not one deployed system.
1 sentence citing source paths or vendor names omitted for publication.
Used in pipelines
Industry tags
- FinanceSource-supported industry
For work involving research materials and financial service operations, use long audio chunk transcription followed by seam alignment and timestamps.
- Media & contentSource-supported industry
For work involving editorial material, programs and reader services, use long audio chunk transcription followed by seam alignment and timestamps.
- HealthcareTransferable use
For organizing consented appointment-service interviews, use long audio chunk transcription followed by seam alignment and timestamps.
Skills pack
Reference pack with offline tests
Download · 63KBSHA-256 789d2fec9b993135491cd119a2778f9e3591a4624848447f75363bdff22b0236
Related components
-
Ad-aware podcast transcript alignment
Audio fingerprint matching + cross-copy time offsets
Working 1 pipeline -
Court calibration and reviewed trajectory reconstruction
Human trajectory annotation correction + calibration reuse and revision audit
Working 1 pipeline -
Video comparison and verified delivery bundles
Before-and-after artifact comparison + review and integrity bundles
Working 2 pipelines -
Podcast and video catalog ingestion
Media metadata normalization + visibility and retention rules
Working 1 pipeline