Background music for podcasts

A podcast really needs three kinds of music: an intro theme that sets the tone, short transitions between segments, and quiet beds that sit under speech. AI generation handles all three well, and generating on a licensed model keeps the whole show cleared for commercial use, which matters the moment you take a sponsor or run ads.

This page covers what each piece is for, the prompts that actually work under a voice, how to mix music so it supports rather than fights the talking, and how to keep one recognisable sound across a whole season.

The three pieces of podcast music

• Intro theme: 10 to 30 seconds, memorable, sets the tone of the show and ideally hints at its genre or mood.
• Transitions: short stings of 2 to 5 seconds that mark a segment change and give the listener a beat to reset.
• Beds: low-energy instrumental loops that play quietly under talking, often used for ad reads, cold opens or storytelling sections.

Most shows reuse the same intro every episode and pull from a small set of beds and stings. That repetition is a feature, not a limitation: it is what makes a podcast sound like itself. So it is worth getting these few pieces right once rather than reaching for something new each week.

Prompts that work under speech

The key for beds is restraint. Ask for low energy, a sparse arrangement and no busy melodies, so the music supports the voice instead of competing with it. Keep it instrumental, and avoid sudden dynamic swings that would force you to keep riding the volume fader. A bed should be almost boring on its own; that is exactly what lets a listener follow the words over the top of it.

Minimal ambient bed, ~70 BPM, soft sustained pads, gentle pulse, sparse arrangement, very low energy, warm and unobtrusive, instrumental, sits under spoken voice.

Confident indie-electronic intro, ~115 BPM, bright synth hook, punchy drums, clean bass, energetic and modern, strong opening, instrumental, around 20 seconds feel.

Short transition sting, ~120 BPM, single bright synth motif, light percussive hit, clean and modern, clear start and stop, 3 seconds feel, instrumental.

Mixing music under a voice

Even a well-chosen bed needs to sit well below the speech in the mix, far quieter than feels right when you audition the music on its own. A common rule of thumb is to bring the bed down until it is clearly subordinate to the voice, then duck it a little further under the words and let it breathe back up in the gaps. If your editor supports it, side-chain or auto-duck the music to the voice track so it steps back automatically whenever someone speaks. Generating low-energy beds in the first place makes all of this far easier, because there is less in the track fighting for the same frequencies as the voice.

A workflow for a new show

1. 1

   Decide the show's tone in a sentence

   Warm and conversational, sharp and newsy, dark and narrative. That sentence drives every prompt you write for the music.

2. 2

   Generate the intro first

   It is the piece listeners hear most. Generate a few takes on a licensed model, pick one, and treat it as the anchor for the rest of the palette.

3. 3

   Generate matching beds and stings

   Reuse the intro's tempo, key feel and instruments in lower-energy form so the beds and transitions clearly belong to the same family.

4. 4

   Save the prompts as a kit

   Keep all the prompts together in your library so producing a new episode is a matter of regenerating from known-good prompts, not starting over.

Keeping a consistent show sound

Save the prompt for your intro and reuse it across episodes, changing only small details when you want variety, for example a slightly different mix for a special episode. A consistent sonic identity is a big part of what makes a show feel produced. If you found a reference theme you love, reverse it into a prompt to capture its tempo and instrumentation, then build your own cleared version from that starting point rather than using the original.