Media Extract / Delivery
Deliver the text—and show where it came from.
Mixed-media delivery needs more than a text file. Users also need to know what was extracted directly, what required recognition, and what needs review.
Multimedia tool · Actual test materials and outputsWhat the next operator needs to know
Run doctor before processing, then inspect text, source method, and status. Reuse successful items and retry failures separately. The workflow depends on macOS and Apple Silicon / MLX for speech recognition, so the environment must be checked before handoff.
01 / Decisions
Hand over the environment, outputs, and retry instructions together.
- Check the environment first
- The doctor command checks decoding, OCR, and model dependencies. The current implementation depends on macOS capabilities; speech recognition uses MLX on Apple Silicon.
- Deliver traceable material
- Text retains page numbers or timestamps, with processing provenance in metadata. Original subtitles, OCR fragments, or segment records are saved alongside the output.
- Support interruption and retry
- Track processing status per file and reuse successful results with --skip-done. Locate failed items individually without restarting the entire batch.
02 / Actual output
Two test materials, two processing paths.
These excerpts come from test fixtures processed on October 26, 2026, not client materials. The Chinese subtitle is translated below for readability; linked outputs retain the actual original text.
- Video with subtitles
- 5-second test video → embedded subtitle extraction → 38 characters in two timestamped segments. Source: subtitle; no speech recognition used for this run.
[0:00] This is the first embedded-subtitle test sentence. [Translation]
View the actual subtitle output (original text) ↗
[0:02] This is subtitle track line two - PDF with a text layer
- 1-page PDF → direct text-layer extraction → 174 characters. Text is organized by page and labeled text-layer.
media-extract text layer test document
View the actual PDF output ↗
The downloadable copies only remove the machine's absolute source path; text and processing provenance remain unchanged.
03 / Verification
Test each branch; do not equate test counts with accuracy.
- Verified paths
- Checked embedded video subtitles, PDF text extraction, and output source/status fields. Scanned-PDF OCR passed with normal local-machine permissions.
- Test coverage
- Full suite rerun on October 26, 2026: 47 passed, zero failures or skips, 15.321 seconds. Covers text processing, naming, PDF, subtitles, speech recognition, and resumable batches. These are fixture-based workflow checks, not general recognition accuracy. View verification notes (Chinese) ↗
- Limitations
- OCR and speech recognition can misread names and require source review. No production usage, accuracy, or time-saving claims are made. No public upload service is provided.