AI import
Audio hides its structure.
A map shows it.
Sound has no shape. An argument that took an hour to build reaches you one sentence at a time, and the structure of it — the main threads, how they connect, which detail belonged to which point — doesn't survive the delivery. Upload the recording: Atlas Red transcribes it for you, then lays that structure out as an editable map you take in at a glance.
How it works
Get the audio file
Most shows publish a direct MP3; most apps have a download button. Your own recordings are already files.
Upload it, choose the depth
Transcription runs first, then the build. Pick overview, balanced or detailed — how much structure you want kept.
See it laid out
The whole shape is visible at once. Rearrange it, prune it, export to any of 14 formats — or hand it to your assistant over MCP.
What actually fits — read this first
One limit is worth knowing before you start rather than after:
20 MB per file, with an outer ceiling of three hours. In practice the file size binds first — 20 MB is roughly half an hour of typical podcast audio, and closer to an hour for a low-bitrate spoken-word recording.
So a full-length show often will not fit whole. Two honest ways round it: import the segment you actually care about, or — if the show publishes a transcript — import that instead, which is cheaper and handles far longer recordings.
Interviews, conference talks, lectures, voice notes and recorded calls are usually comfortably inside the limit. Multi-hour chat shows are the case that needs slicing.
A summary is a shorter sequence.
Structure is a different thing.
Ask most tools to summarize an episode and you get a paragraph — three sentences that could describe a hundred other episodes. It's shorter, and it's still a line of text read start to finish. You can take it in completely and still not know how the ideas stood in relation to each other, because the thing that went missing was never length. It was shape.
A map restores the shape. Everything sits on one surface, in relation to everything else: main threads side by side, supporting detail folded underneath the point it supports, the digression visibly a digression rather than four indistinguishable minutes. You read it the way you read a diagram — spatially, at a glance — instead of reassembling it sentence by sentence.
And because it's editable rather than generated-and-finished, the understanding is something you do: collapse a branch you already knew, move two points together when you notice they were the same point, cut the half that wasn't interesting. What's left is the structure as you understood it — which is the version you'll still recognise next month.
Where the file is
- The show's own website or RSS feed — nearly every podcast serves a direct MP3. This is the most reliable route, and often the only one for a big-name show.
- Overcast, Pocket Casts, Castro and most other apps — a download button that leaves a real file you can share out to Atlas Red.
- Apple Podcasts and Spotify — episodes stay inside the app, so take the file from the show's website instead.
- Your own recordings — voice memos, a recorded interview, Zoom audio, a talk you captured on your phone. These are already files, and they are the case no link-based tool can reach at all.
Accepted formats: mp3, m4a, wav and webm.
Why a file — and not a feed?
Atlas Red doesn't fetch podcast URLs or subscribe to RSS. You download the episode and upload it. That's a real limit, stated plainly — and it buys the thing that matters more. (One link does work: a YouTube video, because YouTube publishes a caption track that can be fetched rather than transcribed.)
A tool that takes a feed URL can only reach what was published. Most recordings worth mapping were not. The customer interview, the internal all-hands, the lecture you recorded with permission, the two-hour strategy call — none of them have a public URL, and none of them ever will. Uploading a file works the same on all of them.
The structure, plus whatever else you want on it
Beyond the shape itself, you can name up to 20 fields before importing — labels of your own choosing, in your own words. There's no fixed set and nothing is required. For a two-person interview you might ask for:
A field is a request, not a guarantee. The AI fills one in only where the recording genuinely supports a value — it's instructed never to invent one — so a point the conversation didn't cover comes back with that field simply empty. That's deliberate, and it's the part worth trusting: an empty field is information, where a confidently wrong one would be a trap.
The other dial is depth. Overview gives a roughly 15-node take on the whole episode, balanced around 40, detailed around 100 — and a sentence of focus narrows the map to the one segment you're actually here for.
What it costs
$1 trial credit
Every new account starts with it — no credit card required, valid 30 days.
AI credit packs
Top up when you like — no subscription, no expiry. Failed jobs are never charged.
Bring your own key Pro
Connect your own account for transcription, building, or both — no markup, ever.
Voice costs more than a document, because transcription is billed by the minute on top of building the map — an hour of audio typically lands in the tens of cents rather than dollars. You never find that out afterwards: the app quotes the job before it runs, and refuses to start one your balance can't cover rather than stopping halfway.
If you bring your own key for only one of the two steps, you're billed credits for the other half — and the quote shows you which half.
Questions people ask
Do I need a transcript, or can I upload the audio?
Upload the audio — Atlas Red transcribes it for you and builds the map from what it heard. mp3, m4a, wav and webm all work. If you already have a transcript or caption file, import that instead: it's cheaper and handles far longer recordings.
How long an episode can it handle?
20 MB per file, with an outer ceiling of three hours — and the file size binds first. That's roughly half an hour of typical podcast audio, closer to an hour for low-bitrate speech. A full-length show usually won't fit whole; import the section you care about, or use the show's published transcript.
Can I paste a podcast link or an RSS feed?
No — Atlas Red doesn't fetch podcast URLs or feeds. You download the audio and upload it. That's deliberate: it's why this works identically on recordings that were never published, like your own interviews and internal talks. YouTube is the one exception: a video link imports directly, since YouTube publishes captions that can be fetched instead of transcribed.
Where do I get the audio file?
Most shows publish a direct MP3 on their site or in their RSS feed, and most podcast apps have a download button. Apple Podcasts and Spotify keep episodes inside the app, so take the file from the show's website instead.
What does a voice import cost?
More than a document — transcription is billed by the minute on top of the build. An hour of audio typically lands in the tens of cents rather than dollars, and the app quotes the exact figure before you spend anything. Every new account starts with $1 of trial credit, valid 30 days, no card.
Can I use my own API key instead of credits?
Yes. With Atlas Red Pro you can connect your own AI account for transcription, for map building, or for both — no markup. Bring a key for only one step and you're billed credits for the other half; the quote shows which.
Will it just transcribe, or actually summarize?
It reads for meaning rather than handing the transcript back. Overview gives a roughly 15-node summary of the episode, detailed around 100 — and a sentence of focus narrows it to one segment.
Does it cope with two people talking?
Yes — interviews and panels are the normal case. You can name a Speaker field and it will attribute points where the recording makes attribution clear. Where it doesn't, the field is left empty rather than guessed at: the AI is instructed never to invent a value it can't support.
Already have the text? Transcripts and caption files. Working from a document? Word, PDF and photos. Already have an outline? Markdown imports free.