Guide Podcasts

AI Podcasts

Scholarly can turn your source material into multi-speaker podcast episodes. Upload PDFs, text files, images, or describe a topic from scratch -- and the platform generates a full audio discussion with transcript, chapters, and citations. A quarterly report becomes a briefing you can listen to on the way to work; a new process document becomes something the team can absorb without reading it.

Creating a Podcast

Start from Home, the New menu, a source's AI actions, or the first-run welcome picker. Podcast creation is then a two-step process.

Step 1 -- Pick your content. Select one or more sources:

  • PDFs -- Upload or choose from your library. For long PDFs, use the page selector to pick exactly which pages the episode should cover.
  • Text files -- Upload .txt, .md, or .csv files.
  • Images -- Upload photos, diagrams, or screenshots.
  • Videos -- Use existing video lectures as source material.
  • Research sessions -- Use completed research reports.
  • Google Drive -- Pick files straight from your Drive, no downloading and re-uploading. See Connections.
  • Link -- Paste a link to a website or an online PDF.
  • Prompt -- Use the Prompt tab to describe any topic and generate a podcast from scratch, no files needed.

Step 2 -- Customize settings. Before generating, you can configure:

  • Style preset -- Choose from Conversational, Exam Prep, or Deep Dive. Each preset changes the structure and depth of the discussion.
  • Length -- Free podcasts are fixed at 2 minutes. Paid plans can use the length slider to set a target duration, from a quick summary to a longer deep dive.
  • Custom instructions -- Add free-text instructions to control the tone, focus areas, or what the speakers should emphasize.
  • Guest voice -- Select a custom voice for the guest speaker, or leave it on Auto and let Scholarly choose distinct voices for you.
  • AI model -- Pick which AI model writes the episode script. Auto uses GPT 6 Luna on every plan, at low reasoning on Free and high reasoning on a paid plan, so an episode left on Auto costs one AI Creation Credit. The picker groups named choices under Frontier Models, Recommended Models, and More Models; anyone can browse, but choosing a specific model is a paid feature. Claude Sonnet 5 and Kimi K3 cost 2 AI Creation Credits, while GPT 6 Astra and Claude Opus 5.5 cost 3. This choice only affects the written transcript -- voices and audio generation are unchanged. See Choosing an AI Model for the full defaults table.
  • Visual companion -- On by default. Generates one chapter slide per chapter alongside the audio. Turn it off for an audio-only episode.
  • Language -- Choose from 70+ supported languages for the generated episode.

Free podcasts are set to 2 minutes and the length control stays locked. After the episode is ready, its persistent Make this podcast longer? banner lets you upgrade and remake it at about 15 minutes in the same podcast — the existing version stays playable until its replacement finishes, and the replacement does not use another creation credit. Paid plans can also set a target length up to 60 minutes before generating. The target guides generation rather than guaranteeing an exact runtime. See Plans and Limits.

Episodes made from a file are titled after the discussion itself, not after the file you uploaded — so a source named Q3_policy_v4_FINAL.pdf comes back as an episode title that says what the conversation is about. Click the title to rename it.

Once you confirm, generation begins. You can leave the page and come back later.

Generation Progress

Podcasts go through several stages:

  1. Pending -- Queued for processing.
  2. Generating transcript -- The AI writes the full script from your sources.
  3. Generating audio -- Text-to-speech records the episode chapter by chapter. On an episode with more than one chapter, chapter 1 becomes playable from this point on, while the rest is still being recorded.
  4. Completed -- The full episode is ready.

You can track the current status on the podcast page at any time.

Start Listening Before It's Finished

If your episode has more than one chapter, you don't have to wait for the whole thing. Chapter 1 becomes playable as soon as it has been recorded, and the page shows which chapters are ready and which are still recording.

Press play and start listening straight away. When the full episode is ready, playback switches over to it and keeps your place — you don't lose your spot and you don't start over.

Voice Quality

Podcast audio is generated with natural, expressive text-to-speech voices that capture tone, pacing, and emphasis. Voices adapt their delivery to the content -- slowing down for technical explanations and picking up energy during engaging moments. All 70+ supported languages use the same high-quality voice generation.

Speakers and Format

Every episode uses a multi-speaker format with a host, a producer, and a guest. Speakers are named after the voice you hear -- for example, Cosmo, Ursa, or Sirius. By default Scholarly chooses distinct voices for the host and producer automatically; if you pick a Guest voice during creation, the guest speaker uses that voice's name instead, while the host and producer stay automatic.

Visual Companion

Every new podcast includes a visual companion by default: one chapter slide alongside the audio, which advances automatically to match the current chapter as you listen. Use the arrows on the slide to jump to the previous or next chapter's visual. You can turn the visual companion off in the customize step for an audio-only episode.

Video Mode

After the episode audio and chapter slides are ready, Scholarly creates a separate 16:9 MP4 in a secure rendering environment. The export matches each slide to its chapter timing and keeps the finished podcast playable even if the video step needs another try. It does not use another AI creation.

Use the Audio / Video switch above the visual companion. Older podcasts with chapter slides can start their MP4 from Video mode. Anyone with access to the podcast can watch or share the video version. MP4 downloads are a paid-plan export; on Free, Download opens upgrade and copying a share link stays free.

Chapters and Citations

Episodes are broken into chapters. Each chapter includes:

  • A title and summary
  • Time markers so you can jump to specific sections
  • Speaker labels showing who is talking
  • Citations that reference the original source material

This makes it easy to find the exact part of the discussion that covers a topic you need to review.

Transcript

A full transcript is available on every completed podcast. Each line is attributed to a specific speaker, so you can read along or search for particular content without scrubbing through audio.

The Player

The player bar is a single row at a fixed height. It does not grow or shrink while you listen: captions, a playback error, or the "resumed where you left off" note float above the bar rather than pushing the play button around mid-episode.

The progress bar is split into one segment per chapter, so you can see the shape of an episode, which chapter you are in, and how far through it you are at a glance. Drag anywhere along it to scrub, or click a chapter's edge to start that chapter.

The chapter list beside it stays quiet until you hover a row. The chapter that is playing is marked with its own indicator and a progress line.

On a phone, the bar keeps the episode title readable instead of clipping it to a few characters.

The three-dot menu offers audio and, when ready, MP4 downloads. Downloads require a paid plan; on Free, Download opens upgrade and copying a share link stays free.

Listening While You Work

Start an episode and it follows you. Open a source, ask a question, browse your library — a mini player stays at the bottom of the screen with play, skip, speed, and the scrubber. Tap the artwork to go back to the full episode.

Podcasts also drive your lock screen, notification shade, and headset buttons, with the title, artwork, and a working scrubber, so you can pause from a headphone tap.

Playback speed and volume follow your account rather than the device, so your laptop and your phone agree. An episode reopens where you stopped — see Continue Where You Left Off.

Interactive Questions

New podcasts come with built-in check-for-understanding questions. While you listen, a quick question appears right after the idea it tests finishes -- the audio never pauses. Answer it to get an explanation grounded in that part of the episode.

Every podcast has an Interactive questions panel under the player. From it you can:

  • Revisit any question from the full list, not just the one that just played.
  • See your Best and Last scores, plus a full history of past attempts.
  • Retake the questions as many times as you like -- only completed attempts count toward your Best.

Prefer to just listen? Turn the toggle off in the questions panel at any time -- the setting is saved to your account, so future episodes follow the same preference until you change it.

Captions

Every completed podcast comes with synced captions, on every plan including Free. The player has a dedicated CC button: click it to switch captions on, click it again to hide them. Captions are off the first time you open a podcast — turn them on once and the player remembers your choice for the next episode.

Captions carry the same speaker labels as the transcript, so you always know who is talking even with the sound off. The names follow the voices in the episode: Cosmo, Ursa, Sirius, or whichever voice you picked for the guest. An episode only labels the speakers actually in it, so a solo episode names a single voice.

Captions are written from the same script that drives the narration rather than transcribed from the finished audio, so they match the spoken words exactly and land on the right moment. They are generated in the language you created the episode in — a French episode comes with French captions, with no separate setup. Regenerate the episode and the captions regenerate with it.

If an Episode Can't Be Created

A failed episode says what went wrong -- the content was declined, a source couldn't be read, a source was too large, the sources didn't hold enough to work from, or the script or narration step broke -- and offers only the fixes that apply to that reason: Try again, Try another model, Pick fewer pages, Replace the source, or Adjust the sources.

When the cause lives in the source itself, a bare Try again is deliberately not offered, because replaying the same sources would fail the same way. Your sources and settings are restored into the panel, so you are not rebuilding the request from scratch, and a creation that fails returns its AI Creation Credit automatically. See When a Creation Fails for every reason and fix.

Renaming

Click the podcast title to rename it. This is the same control that lets you replace an auto-generated episode title with your own.

Sharing

Share any podcast with a link. Recipients can listen to the episode, read the transcript, and browse chapters without needing to create an account.

Creating from AI Chat

You can create podcasts directly from the AI chat. Just say something like "make a podcast about this vendor contract" or "turn this into a podcast" and the AI handles the rest. When the podcast is ready, it plays inline in the conversation -- no need to navigate away.

Creating from an Upload

You can also create a podcast when you upload files. Add your files, then choose Create Podcast from the primary create tiles. A customization panel lets you pick a style preset, guest voice, language, and custom instructions before generating.

AI Chat

Every completed podcast has a dedicated AI chat panel. Ask questions about the episode content, request clarification on a topic, or explore related ideas. The AI uses the podcast transcript and source materials to give grounded answers.

Was this helpful?
PreviousVideo LecturesNextResearch