AI summary
AI that fits what you record,
built in, not bolted on.
A generic summary treats a Reel, a sales call, and a lecture the same way. Pepys doesn’t. Every transcript can be read through a framework shaped for what it actually is, with a layout to match. Six of them, all live today.
Six frameworks. One upload.
Pick the lens that matches your recording, or let the default catch it. Each one is a real analysis option in the tool, with its own purpose-built layout, shown live below.
Short-form video
Hook, CTA & retention structure for TikToks / Reels / Shorts.
Creators live and die by the hook. This framework names your content type, pulls the exact opening line, and explains why it stops the scroll, then maps the retention beats from tension to payoff. You also get the core message, the on-screen CTA, topic tags, and ready hashtags, all read straight from what you actually said.
General
TL;DR, key points & a title – for anything.
The default for anything you upload. It gives you a clean title, a TL;DR you can paste into a doc, and the handful of key points that carry the whole recording, so a long file becomes something you can skim in seconds.
Podcast / long-form
Show notes, key topics, pull-quotes & guests.
Built for long-form episodes. It writes the summary, lists who is on the mic, and surfaces the key topics, the pull-quotes worth clipping, and a show-notes outline, so the post-production writing is mostly done before you open your editor.
Meeting / sales call
Decisions, action items, next steps & open questions.
Made for calls where the follow-through is the point. It separates the decisions that were made from the action items, each with an owner, then lays out next steps and the open questions still hanging, so nobody has to rewatch the recording to remember who owns what.
Lecture / study notes
Key concepts, an outline, takeaways & practice questions.
Turns a recorded class into something you can study from. It defines the key concepts in plain terms, lays out the lecture as an outline, distills the takeaways, and writes practice questions, so revision starts from real notes instead of a wall of text.
Interview
Themes, notable quotes & a Q&A breakdown.
For conversations you need to quote accurately. It pulls the recurring themes, lifts the notable quotes verbatim, and reorganizes the talk into a clean question-and-answer breakdown, so finding the line you remember takes a glance, not a scrub.
The hook
“We launched with everything... and ninety-five percent of our active users only ever touched one thing.”
Core message
Subtraction is underrated: when the data showed almost everyone used only the weekly meal planner, cutting three quarters of the app roughly quadrupled retention – and later killing the free tier tripled revenue. Trust what users do, not what you hoped they'd do.
Retention structure
- 1
Confession hook: launched with a dozen features, but ~95% of users only used one of them
- 2
Tension: eight months and most of the runway spent building things almost nobody opened
- 3
Turn: cut three quarters of the app despite team pushback – 'subtraction is underrated'
- 4
Proof: retention roughly quadrupled the moment the app got smaller
- 5
Escalation: killed the free tier – signups fell ~30% but revenue tripled
- 6
Payoff line: 'data beats ego' – find the one feature your best users can't live without and delete the rest
Topics
Suggested hashtags
More than a generic summary.
The same paragraph-shaped summary for everything misses what each recording is for. This is built to read your file the way it was meant to be read.
Shaped to the format
A sales call and a lecture aren't the same, so they aren't read the same way. Each framework pulls what matters for that kind of recording, not a one-size blurb.
A layout built for it
Show notes read like show notes; meeting notes surface the decisions and action items with owners. The structure matches the job, not a wall of text.
Runs on every transcript
The analysis is drafted automatically the moment your audio is transcribed. No extra step, and no pasting the transcript into a separate AI tool.
Yours to edit and export
Every draft is a starting point, not a black box. Fix anything inline, then export to DOCX, PDF, or plain text in a click.
Better together
Then ask it anything.
The summary gives you the gist. When you need a specific answer, chat with the same transcript and get it back grounded in the audio, cited to the second.
Common questions
- What does the AI summary actually do?
- When your audio or video is transcribed, Pepys reads the transcript and drafts a structured summary shaped for what you recorded, with a layout to match. It runs automatically on every transcript, so the notes are ready when the transcript is.
- What kinds of summaries can it make?
- Six frameworks, all live today: hooks and retention beats for short-form video, show notes for podcasts, decisions and action items for meetings, study notes for lectures, themes and quotes for interviews, and a clean general summary for anything else.
- Does it pick the right one automatically?
- Yes. Pepys defaults to the framework that fits the source it detects, and you can switch to any other lens with one click and re-run it on the same transcript.
- Is the summary accurate, or does it make things up?
- It works only from your transcript. Quotes are lifted verbatim from what was said, and anything it gets wrong you can correct inline. It's a draft grounded in the recording, not a guess.
- Can I edit and export the result?
- Yes. Edit any part of the analysis inline, then export to DOCX, PDF, or plain text. You can also ask the same transcript follow-up questions with AI chat.
- What languages does it work in?
- It auto-detects across 99+ languages and writes the analysis in the transcript's language by default. You can also set a fixed output language if you'd rather always get it in one.
Same upload. Different intelligence.
Drop a file or paste a link, pick the framework that fits, and read your recording the way it was meant to be read.
Transcribe something