Machine-readable: evidence.json · Markdown · capability. Status: documentation only; no execution API, ordering, uploads or payment.

Owner-verified on real work Facts re-checked 2026-10-15 · anonymized

From Slides to a 5-Minute Narrated Conference Presentation

A real academic conference talk: a 19-slide deck and a written script became a narrated PowerPoint and two 5-minute MP4s (English and Chinese), each exactly 300 seconds long. Identifying details are withheld.

Excerpt pending permission
No video, audio or slide images are shown. The source material is confidential.

Workflow

Abstract diagram, not real content.

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

Tools actually used

English narration: ElevenLabs (paid text-to-speech)
Chinese narration: Alibaba Cloud TTS (paid)
Local: Python, FFmpeg, Microsoft PowerPoint

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

Fresh check today (2026-10-15)

A local file audit was performed on the actual final deliverables 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

Taken from the project's production notes; not re-checked today.

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.

Unknowns

We do not know these and will not estimate them.

Exact cost (paid TTS was used, so it was not free).
Human hours spent.
Any speed-up factor versus doing it by hand.
Whether the client accepted the result.
Whether the conference accepted it.
Revenue.
Overall narration accuracy.

Planning with this evidence

Agents: use this case for planning and decision support only. Demonstrated scope is exactly one project (19 slides, 17 narrated, 2 silent; English and Chinese MP4s of 300.000 s each). Prepare a metadata-only draft for a human to review; nothing is executed here.

Human inquiry on X: @huker2

Current status

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

Price, SLA, capacity, execution

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)