# From Slides to a 5-Minute Narrated Conference Presentation

> Owner-verified, anonymized case. Facts are split into a fresh local file audit, production records, and unknowns. Source material is private; this is a summary, not public reproducible proof.

Machine-readable: [evidence.json](https://workflows.dubspeak.com/cases/slides-to-narrated-presentation/evidence.json) · [capability](https://workflows.dubspeak.com/capabilities/narrated-presentation.json)

## Status
- Documentation and discovery files are available.
- llms.txt, Markdown and JSON documents can be read.
- There is no job execution API.
- Automatic ordering is not implemented.
- Uploads are not implemented.
- Payment is not implemented.
- Automated jobs are not accepted.
- Email inquiry (version 1) is available: one JSON request per email to dubspeak+inquiry@mail.grokbot.com, reviewed by a human.
- Manual contact is also possible via the public X account @huker2 (https://x.com/huker2).

## Workflow
- Deck + script
- Page mapping
- Pronunciation review
- Per-slide TTS
- Audio + timeline
- PPTX + MP4
- Exact-duration checks + human review

## Tools actually used (production records)
- ElevenLabs — English narration (paid text-to-speech)
- Alibaba Cloud TTS — Chinese narration (paid text-to-speech)
- Python — local scripting
- FFmpeg — local audio/video assembly
- Microsoft PowerPoint — local deck timing and export

Not free, not fully automatic. Paid TTS was used. Human review and multiple revisions were part of the production process.

## Fresh local file audit (2026-10-15)
A local file audit was performed on the actual final deliverables by the owner's local assistant, on the owner's behalf.
- Final PPTX: 19 slides; slide transition timings sum to 300,000 ms; ZIP integrity check OK.
- English and Chinese MP4: H.264 video / AAC audio, 1920×1080, 30 fps, 9,000 frames.
- Video stream, audio stream and container duration are each exactly 300.000 s in both MP4s.
- File hashes match the existing final QA record.

## From production records (not re-checked in the audit)
- 17 slides carry narration; 2 slides are silent.
- A 1-second pause precedes narration on each narrated slide.
- Embedded media in the deck was kept.
- Pronunciation fixes, a logo fix and timing QA were done during production.

## Demonstrated scope (no guarantees beyond it)
- 1 project; 19 slides (17 narrated, 2 silent)
- English MP4 and Chinese MP4, each 300.000 s, 1920×1080 at 30 fps, 9,000 frames

## Unknowns
- Exact cost (paid TTS was used, so it was not free): unknown (null)
- Human hours spent: unknown (null)
- Overall narration accuracy: unknown (null)
- Whether the client accepted the result: unknown (null)
- Whether the conference accepted it: unknown (null)
- Revenue: unknown (null)
- Any speed-up factor versus doing it by hand: unknown (null)
- Price: unknown (null)
- SLA: unknown (null)
- Capacity: unknown (null)
- Execution: not available (no endpoint)

## Limits
- Exactly one owner-verified project is documented. Nothing here guarantees support for other page counts, durations or languages.
- The raw slides, script, voice and output files are private and are not published.
- The metadata is a summary, not public reproducible proof. A third party cannot re-run or re-check it from this site.
- The fresh file audit was performed by the owner's local assistant on the owner's behalf; this website did not run the project.
- Production-record facts were not re-checked in the fresh audit.
- Hashes of the private files are not published. Any hash in a release manifest is a document integrity hash only and proves nothing about an outcome.

Media: Excerpt pending permission. Version 1.0.0 · 2026-10-15