How to Make an Actual Play Podcast With AI Voices

To make an actual play podcast with AI voices, you write or transcribe the session into a script, assign a distinct voice to each player character and the game master, generate the dialogue as audio, layer in sound effects and music for scenes and combat, then export one finished episode file. With AudioProducer you produce that episode audio and download the MP3. You then upload it to your own podcast host, which builds the feed and lists the show. We make the file; we do not host, distribute, or publish your podcast for you.

This route suits a solo creator or a small group that wants a polished, consistent show without booking a recording session, chasing everyone's microphone quality, or editing hours of crosstalk. Below is how the pieces fit together.

What an actual play podcast is

An actual play show follows characters through a tabletop roleplaying game: a party of adventurers, a game master narrating the world, dice deciding what happens, and the story that emerges from those choices. Popular shows record a live group playing, but the format itself is just characters, a narrator, and events unfolding scene by scene. That structure maps cleanly onto scripted or lightly-scripted audio, which is exactly what AI voices are good at producing.

Worth naming the difference up front: an actual play keeps the shape of a game session, including the game master describing scenes and the back-and-forth of players reacting. That sets it apart from a fully scripted audio drama built from a campaign, where the game framing usually disappears and you are left with a straight narrative. It also differs from turning a rulebook into narration, which is closer to a sourcebook read aloud. Decide which one you are making before you write a line, because it changes how much game texture you keep.

Script or capture: two ways to build the episode

There are two common starting points. The first is to write the session as a script from scratch: you plan the encounter, draft each character's lines and the GM's narration, and decide the beats in advance. This gives you the tightest, most listenable result because every pause and reveal is intentional.

The second is to run or play a real session, capture what happened, and shape that transcript into a script. You keep the surprises that make actual play fun, then trim the dead air, cut the rules debates, and clean up the lines so they read well as audio. Either way you end up with a script: a list of who says what, in order, with narration marked separately from character dialogue.

Mark your script clearly. Put the GM narration in its own blocks, tag each character line with the speaker, and note where a scene changes or where you want a sound cue. That structure is what lets you assign voices quickly in the next step.

Cast a voice for every character and the GM

The thing that makes an actual play listenable is that you can tell who is talking without a name tag every line. Give the game master one steady voice that carries narration and world description, then give each player character a voice that stays consistent from episode to episode. A gruff fighter, a quick-talking rogue, a measured wizard: the contrast does the work.

Build a simple voice chart before you generate anything: character name, the voice you picked, and a one-line note on their delivery. Keeping that chart means episode nine sounds like episode one. For the mechanics of pulling apart similar-sounding characters so a crowded party stays clear, the same approach we cover for giving each character a different voice applies directly here. If you want to clone a real voice for a recurring NPC or a co-host, use only a voice you own or have permission to use.

Add sound effects and music for scenes and dice moments

Actual play lives on atmosphere. A tavern hum under the party's downtime, the scrape of a door in a dungeon, a swell of music when initiative is rolled: these cues tell the listener where they are and when the stakes just changed. You do not need a wall of effects. A handful of well-placed sounds per scene reads as production value; a constant barrage buries the dialogue.

Plan cues scene by scene as you script. Note an ambience bed for each location, a short sting for combat or a critical hit, and a music shift for the emotional turns. The same workflow we describe for adding sound effects and music to long-form audio carries over: keep beds low under the voices, let effects punctuate rather than distract, and mix so the dialogue is always the clearest thing in the room.

Export the episode and publish it to your host

Once the dialogue, effects, and music are set, you generate and export the full episode as a single MP3 and download it. That file is the deliverable from our side. To get it in front of listeners, you upload it to a podcast host (the service that stores your audio and generates your RSS feed). The host is what actually lists the show on the apps; we produce the episode, and you publish it where you already work.

A few production habits pay off across a season: open with a short cold-open scene before your intro, keep episode length consistent so listeners know what to expect, and give the show a strong front door. If you want the opening minute to hook people, the guidance for making a podcast trailer works for episode intros too. Disclose that the show uses AI voices where your platform or audience expects it; being upfront costs nothing and avoids surprises.

Getting started

You can try the whole flow before committing anything: the free plan gives you 1,200 words with no card, and paid plans start from $39.99 per month. Draft a short scene, cast two or three voices, drop in a single ambience bed, and export it. Hearing a two-minute clip of your own party talking is usually enough to tell you whether the format is worth a full season.

Frequently asked questions

Can I mix AI voices with my own recorded narration? Yes. Many creators voice the game master themselves and use AI voices for the player characters, or the reverse. You assemble the lines into one script, generate the AI parts, and combine them into the exported episode.

Does AudioProducer publish my podcast to Spotify or Apple? No. We produce and export the episode audio file. A podcast host builds your RSS feed and lists the show on the apps; you upload the exported MP3 there yourself.

How long should an actual play episode be? There is no fixed rule, but consistency helps more than length. Pick a target that fits your scripting pace, whether that is twenty minutes or an hour, and keep episodes in a similar range so listeners know what to expect.