Sample · Podcast Pack
What a package contains
Synthetic sample · fictional business, invented figures, not a real client or quote
From the bundled synthetic episode “Why Pipeline Reviews Fail”, about 4 minutes, with a fictional host and guest voice. Every quote below is copied from the supplied transcript; nothing is generated.
What you supply (current limits)
- One episode file of up to 96 MB and 60 minutes (WAV, MP3 or MP4). A 60-minute uncompressed WAV is about 660 MB, so compress it first.
- A timestamped transcript. There is no automatic transcription at present.
- Your show name, host and description, used in the package.
Files returned after operator review
audio.mp3 processed towards −16 LUFS; the delivered level is measured notes.md chapters, verbatim key points, loudness table chapters.json chapter start times and titles captions.srt captions transcript.srt / transcript.json thumbnail.txt text-overlay suggestions (not a designed image) description.txt clip-01-hook.mp4 vertical 720×1280 clip with burned-in captions clip-02-insight.mp4 manifest.json SHA-256 checksum for every file README.txt one .zip of everything
The current setup produces two clips per episode. Very short sources may give fewer.
Chapters (from the transcript)
- 00:00 Welcome back to Signal and Scale. I'm Dana, and today we are…
- 01:19 The first month was uncomfortable. Reps felt watched, and…
- 02:46 It is, and for those accounts we still send the long version…
Chapter titles start as the opening words of each section; the operator edits them before delivery.
Key points (verbatim excerpts)
- (00:35) “Stage age, which is the median days a deal has been in its current stage.”
- (01:37) “By the third month, the median stage age in the proposal stage dropped from nineteen days to eleven days.”
- (02:23) “The new one is three pages: the problem in the buyer's words, the outcome with a date, and the three decisions the buyer needs to make to start.”
Loudness, honestly
Audio is normalised towards −16 LUFS with a −1.5 dBTP ceiling. The delivered file is measured, and a miss is flagged for review rather than hidden. This sample measured −16.8 LUFS after processing and was flagged. Processing is loudness normalisation and optional edge trimming only: no dialogue editing or noise repair.