The short version: most "sped up" podcasts aren't running faster, they've had their pauses stripped out by editing software, been recorded and imported at mismatched sample rates, or been quietly time-compressed by a few percent to hit a target runtime. Fix the pause length, match your sample rates, and stop nudging speed to save two minutes, and the problem disappears.
It's rarely the playback speed setting
I get asked this a lot, usually by clients who've just launched a show and someone in the comments has said "sounds like you're on fast forward." First thing they check is whether their podcast app defaulted to 1.2x. It hadn't. That's almost never the answer once you've confirmed the app is set to normal speed.
What's happening is one of three things, and they get lumped together under "sped up" because the effect on the ear is similar: a slightly higher pitch, a breathless cadence, sentences that seem to bump into each other with no room to land.
Cause one: sample rate mismatch
This is the technical one and it's more common than people think, especially with remote interviews. If you record at 44.1kHz but your editing project is set to 48kHz (or the other way round), the software plays the file back at the wrong speed to fit the timeline. The math is exact: 48 divided by 44.1 is 1.088, so a file recorded at 44.1kHz and dropped into a 48kHz project plays back roughly 8.8% faster, with a matching pitch rise. It's small enough that most listeners can't name it, but they feel it. It sounds like the host has had three coffees.
This happens constantly with Zoom recordings. Zoom's default audio input on some systems is 44.1kHz, but a lot of editing templates (Audition, Hindenburg, Descript projects built from a podcast template) default to 48kHz. Nobody checks. The episode gets exported, it sounds slightly wrong, and everyone blames the guest's internet connection instead of the settings panel.
Cause two: pauses cut too tight
This is the bigger culprit and it's a workflow choice, not an accident. Tools like Descript, Auphonic and even Audacity's noise gate have a "remove silence" or "remove filler words" function with a threshold you set, usually in milliseconds. If that threshold is set to strip anything under 150 to 200 milliseconds, you're removing the natural micro-pauses people take between clauses, not just the dead air between sentences. The words are all still there at their normal individual speed, but they're packed together with no breathing room. Your brain reads that compressed cadence as speed even though the actual phoneme-level pace hasn't changed.
I worked with a client last year, a two-person business podcast, where the editor (a freelancer they'd hired off a marketplace) had set Descript's silence removal to its most aggressive preset because he thought tighter editing sounded more "professional." The finished episodes ran about 15% shorter than the raw recordings. Both hosts sounded like auction callers. We didn't touch the speed at all. We just changed the pause floor to 400 milliseconds and re-exported, and the exact same recording suddenly sounded calm and normal. Nothing about the actual voices changed. Only the gaps between them did.
Cause three: deliberate time compression
Here's the part most guides on this topic skip over, because it makes producers look bad: a lot of shows are quietly sped up on purpose, and it's an accepted practice at the network and agency level that almost nobody discloses to listeners. If a 46 minute recording needs to fit a 40 minute ad-supported slot, the easy fix isn't a harder edit, it's applying 2 to 5% time compression across the whole file in a tool like Adobe Audition or Auphonic's "speed correction," which shortens runtime and raises pitch slightly using algorithmic time-stretching rather than a simple speed-up, so it's less obviously chipmunk-voiced than old-school pitch shifting. Under about 3% most listeners can't identify what's wrong, they just say the host "sounds off" or "sounds hyper today." Above 5% it becomes noticeable even to casual ears. This is common enough that if you've ever felt a network-produced show sounds subtly different from an indie one recorded the same way, this is often why. It's a legitimate editorial choice for hitting ad breaks, but treating it as invisible to the audience is the bit nobody wants to admit out loud.
How to check what's going on with your own show
Before you change anything, work out which of the three it is.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
- Open your raw, unedited recording and your final published episode side by side.
- Find a phrase both files share, ideally near the start before heavy editing kicks in, and time how long it takes to say in each version using the waveform view.
- If the phrase takes noticeably less time in the final file but the pitch sounds slightly higher, that's compression or a sample rate mismatch, not editing.
- If the phrase takes about the same time in both but the overall episode is much shorter, that's pause removal doing the work, and your fix is in the silence settings, not the speed settings.
- Check your recording software's input sample rate and your editing project's sample rate. They should match exactly. 48kHz is the safer default for most modern editing tools and matches Zoom, Riverside and Squadcast's usual recommended settings.
The fix, step by step
Once you know the cause, the fix is boring, which is good, because boring fixes are the ones that stick.
- Match sample rates everywhere. Set your recording app (Zoom, Riverside, whatever you use) to 48kHz and your editing project to 48kHz. Check this once and it stops being a problem forever.
- Set silence removal to no less than 300 to 400 milliseconds. Descript, Auphonic and Hindenburg all let you set this manually. The default presets are often built for tightening corporate video, not conversational speech, and they'll cut too close.
- Turn off automatic speed correction entirely unless you're doing it deliberately and sparingly. If you need to trim runtime for ad slots, cap it at 1 to 2%, and do the rest of the trim by cutting actual dead content (a rambling tangent, a repeated point) rather than shaving speed off everything.
- Use a slower limiter release, not a brickwall one. Fast, aggressive limiting causes audible "pumping," where the volume snaps back up right after loud syllables, and that pumping reads to the ear as rushed pacing even when the actual timing is untouched.
- Target -16 LUFS for stereo distribution or -19 LUFS for mono, which is the loudness standard Spotify and Apple Podcasts both recommend. LUFS (Loudness Units Full Scale) is a standardised loudness measurement, and you can read more about how it works on Wikipedia's page on loudness units if you want the technical detail. Hitting a sane, consistent loudness target reduces how hard you need to limit in the first place, which reduces the pumping that makes shows sound frantic.
What good pacing sounds like
A well-paced podcast still has "dead" moments. A host pausing half a second before answering a hard question isn't a mistake to be edited out, it's the sound of someone thinking, and it's part of what makes a conversation feel real rather than like a script being read at speed. The over-edited, over-compressed version of a podcast is technically "cleaner" by every metric an editor can point to on a waveform, and it's worse to listen to for 40 minutes. I'd rather hear three "ums" and a proper pause than a guest who sounds like they're reading an auctioneer's script because every gap under half a second got vaporised in post.
If you outsource editing, and most growing shows eventually do, ask the editor directly what their silence removal threshold is and what sample rate they work in. Those two questions will tell you more about how your show is going to sound than almost anything else in the brief.
Frequently asked questions
Why does my podcast sound sped up even though I checked the playback speed?
Playback speed settings in the app are almost never the cause. It's far more likely that your editing software's silence or pause removal is set too aggressively, or your recording and editing sample rates don't match, both of which create the perception of speed without changing word-by-word pacing.
What sample rate should I record my podcast at to avoid this?
Record at 48kHz and set your editing project to 48kHz as well. This is the standard most modern recording tools like Zoom, Riverside and Squadcast default to, and keeping recording and editing rates identical avoids the pitch and speed shift that happens when they mismatch.
Is it ever okay to deliberately speed up a podcast episode?
Small time compression, up to around 2%, to fit an ad-supported runtime is common practice and mostly undetectable to listeners. Anything beyond about 3 to 5% starts to sound noticeably off, with hosts coming across as breathless or artificially energetic, so it's worth using sparingly and combining with real content trims rather than relying on it alone.
How short should pauses be between sentences in a podcast edit?
Keep natural pauses at 300 to 400 milliseconds at minimum when using automatic silence removal tools. Cutting below that strips out the natural rhythm of speech and creates the compressed, rushed cadence listeners describe as "sped up," even though the actual word speed hasn't changed at all.
