AI music for podcasts: intros, beds and stingers you actually own the files to
A podcast needs less music than people assume and more consistency than they expect. Three pieces do almost everything: a theme that opens every episode, a bed that sits under narration, and a short stinger that marks a section break. Generate those three once, use them for a year, and your show sounds like a show.
This covers how to generate each one, what to ask for, and the licensing question you should settle before you publish.
The three pieces
The theme. Twenty to forty seconds, with a recognisable hook in the first five. It plays at the top of every episode, so it is the single most-heard piece of audio you will ever make. Spend the most effort here.
The bed. Two to four minutes of low-event instrumental that sits under speech without competing with it. No sudden dynamics, no strong melody in the vocal range, nothing that resolves in a way that draws attention. A good bed is boring on its own; that is the job.
The stinger. Two to four seconds. A short motif, ideally derived from the theme, that punctuates a transition. Having one makes an edit sound deliberate.
Prompting for each
The Music Studio generates full original tracks — instrumental or with vocals, across 35 languages — at 25 credits per track. The prompts differ sharply by piece.
Theme prompt pattern: genre, instrumentation, tempo, mood, and the word intro.
Warm analog synth intro, mid-tempo, curious and optimistic, arpeggiated lead over soft pads, builds to a clean stop
The words builds to a clean stop matter more than they look. A theme that fades is hard to edit against speech; a theme that ends decisively gives you a hard in-point for the first line of narration.
Bed prompt pattern: ask for restraint explicitly.
Minimal ambient instrumental bed, slow, no melody, soft sustained pads and light texture, unobtrusive, consistent throughout, no dynamic swells
Every one of those negatives is doing work. Generators default to interesting, and interesting is exactly wrong under speech.
Stinger prompt pattern: short and specific.
Very short three-note synth motif, bright, clean ending, no reverb tail
If you want the stinger to relate to the theme, describe the theme's instrumentation and mood in the stinger prompt. Generation is not aware of your other tracks, so continuity comes from the prompt.
Loudness: the part most shows get wrong
Music under speech should sit far lower than instinct suggests. A bed that feels "slightly too quiet" in isolation is usually about right under narration, and a bed you can comfortably hear is almost always too loud.
Two practical habits:
- Mix the bed while listening on the worst speakers you own — phone speaker, laptop, one earbud. If the words survive there, they survive everywhere.
- Duck the bed under speech rather than riding a single level. Even a crude manual dip at the start of each speech segment beats a flat bed.
Set the theme's peak level and the bed's level once, write them down, and reuse those numbers every episode. Consistency across episodes matters more than perfection in one.
Length and structure
Generated tracks come out as complete pieces rather than loops, which is convenient for a theme and inconvenient for a bed — a four-minute bed under a fifty-minute episode needs to repeat.
Two approaches:
- Generate several beds in the same prompt style and rotate them across the episode. Variation you did not plan reads as production value.
- Loop one bed on the timeline, cutting at a point where the texture is steady. Ambient beds with no strong rhythm loop almost invisibly; anything with a beat needs the cut to land on the beat. The BPM tapper will tell you the tempo if you need to count bars.
For the theme, generate three or four candidates before choosing. It costs 25 credits each and you are picking something you will hear several hundred times.
The licensing question, answered honestly
This is where podcasters get nervous, and the honest answer has two parts.
What we provide: the tracks you generate are original audio produced for you, and you can use them in your podcast. Downloads are branded in the filename; there is no visible watermark and no audible tag over the music.
What nobody can promise you: the legal status of AI-generated music is still developing, and it varies by jurisdiction. Anyone offering you an unconditional copyright guarantee on generated audio is overstating what is currently knowable. Our AI music licensing explainer sets out the actual position, and can you use AI music on YouTube covers how platform systems treat it in practice.
The practical implication for a podcast: generated music avoids the specific, common problem of using a commercial recording you have no rights to. It does not make you immune to every conceivable future claim, and you should treat any tool that says otherwise with suspicion.
Disclosure
If your show is distributed in a jurisdiction with AI-labelling rules, music may fall inside them. Exports carry C2PA content credentials as provenance metadata, which is the machine-readable half of that. Our guide to AI content labelling under the EU AI Act covers where explicit disclosure is required and where it is not.
A one-session setup
- Generate four theme candidates. Pick one. (25 credits each.)
- Generate two beds in a deliberately restrained style.
- Generate one stinger echoing the theme's instrumentation.
- Set levels once against a real episode.
- Save the prompts. When you want a seasonal variation, you regenerate from the same prompt rather than starting over.
Five tracks is 25 credits multiplied by five, well inside the 150 one-time credits a free account starts with — enough to finish the whole setup before deciding whether to pay for anything.
If music is the part you are working on, these go further: Music for Twitch streams, AI music for streamers, Background music for videos, How to add music to a video, and Free vocal removers compared.
Frequently Asked Questions
Can I use AI-generated music in a monetised podcast?
Yes, that is the intended use. The caveat worth understanding is that AI music licensing is an evolving area rather than a settled one, so treat any absolute guarantee sceptically — including from vendors. Our licensing explainer covers what is actually known.
How long should a podcast intro be?
Twenty to forty seconds for the music, with the recognisable hook inside the first five. Listeners skip long intros, and the recognisable part is what earns the seconds you do get. Generate the theme with a clean ending rather than a fade so you have a hard edit point.
What makes good background music for narration?
Restraint. No strong melody in the vocal frequency range, no dynamic swells, no resolution that pulls attention. Prompt for it explicitly — ask for "no melody" and "unobtrusive" — because generators default to interesting music, and interesting competes with your voice.
Can I get instrumental-only tracks?
Yes. Ask for instrumental in the prompt and do not supply lyrics. If you want a vocal theme, the Music Studio handles vocals across 35 languages, and our songwriting guide covers writing the words.
Do I need a different track for every episode?
No, and you should not want one. Recognition is the point of a theme — changing it costs you the association you are trying to build. Rotate beds if you like, keep the theme fixed, and save the seasonal variation for something worth marking.
Build the set once
Open the Music Studio, generate a theme, a bed and a stinger in one sitting, and set your levels against a real episode. A free account includes 150 one-time credits with no card, which covers the whole set with room to reject a few candidates — and credits do not expire, so the ones you do not spend are there next month.
Produce original AI music with custom vocals for your videos and streams
Create complete custom tracks with vocals in 35 languages, extract stems, generate cover art, and export finished songs.