The short version. Riverside fixes the problem you cannot fix later, by recording each guest locally instead of capturing the call. Descript is the edit, because a conversation is faster to cut as text than as waveform. Auphonic is the one that makes every episode sound like the same show. Castmagic turns a finished episode into the show notes and clips you keep not writing. ElevenLabs is for ad reads, pickups and other languages. And its free tier grants no commercial licence, which matters more here than anywhere else on the list.
Prices below were read off each vendor’s own pricing page on 15 August 2026. Auphonic is the exception and we say why in its section: its page sells tiers by processing hours rather than a headline figure, so we have quoted the hours.
What we weighted, and why the edit isn’t it
Podcasting has an unusual shape as a production problem. Almost everything that determines whether the episode is good happens before you open an editor, and almost everything that determines whether anyone hears it happens after you close one. The middle, the edit, is the part people buy tools for, and it is the part that matters least.
So we weighted the ends:
- Does it protect the recording? Bad audio is the only defect in this medium that listeners will not forgive, and it is the only one you cannot fix in post.
- Does it survive a guest? Your co-host will install anything. A guest gives you one take, on their laptop, with whatever bandwidth the hotel has.
- Does it reduce the work that happens after the edit? Show notes, chapters, clips and descriptions are where independent shows quietly die, because they are unpaid and endless.
- Does the licence cover a sponsored episode? The moment money is involved, “free tier” and “commercially licensed” stop being the same question.
We also assumed you are not making a video podcast primarily. If you are, the video tooling dominates the decision and this is the wrong list.
1. Riverside: best for the problem you cannot fix later
Free tier with 2 hours of recording, one-off. Pro $29/month, or $24/month billed annually ($288/year), for 15 recording hours a month. Grow $39/month, or $34 annually, for 20 hours. Verified 15 August 2026.
The single highest-leverage purchase in podcasting, and it is not an editing tool. Riverside records each participant locally, at full quality, and uploads the files afterwards. So the connection quality during the call determines how the conversation felt, not how the recording sounds. A call recorder gives you a permanent artefact of your guest’s wifi.
This matters because of the asymmetry in the medium. You can rescue a mediocre edit. You can rewrite bad show notes. You cannot un-compress a guest’s audio, and you will not get the interview again. Most guests give you one hour, once, and a re-record request is an admission you will not want to make.
Track the recording-hour allowance rather than the monthly price. It counts every participant-hour you capture, which is what surprises people: a 90-minute session with two guests is not 90 minutes against your allowance.
Who it’s wrong for: solo podcasters recording into a local interface. If nobody joins you remotely, you are paying for the one feature that justifies the product. Record into whatever DAW you like and spend the money on Auphonic instead. Also anyone whose guests reliably refuse to open a browser tab and install nothing; a phone call you record badly still beats an interview you never got.
2. Descript: best because a conversation edits faster as text
Free tier with 60 media minutes a month. Paid tiers per user, per month, billed annually: Hobbyist $16 (600 media minutes), Creator $24 (1,800), Business $50 (2,400). Month-to-month costs materially more; Descript advertises up to 35% off for annual. Verified 15 August 2026.
Podcasting is the use case Descript was built around, and it is the one where the transcript-first idea is unambiguously correct. Cutting a tangent is selecting a paragraph. Removing filler words is a checkbox rather than an afternoon. Finding the good ninety seconds in a two-hour conversation is a text search, not a scrub.
Studio Sound is the feature that quietly justifies the subscription for most independent shows, because the realistic alternative is not a better plugin. It is nobody doing anything about the guest who sounded like a tunnel.
The constraint is media minutes, and they are counted on what you upload rather than what you publish. A weekly show recorded as separate tracks for host and guest consumes allowance at twice the length of the episode before you have cut anything, which is why Hobbyist’s 600 minutes goes faster than the number suggests.
Who it’s wrong for: heavily produced narrative shows with music beds, sound design and layered ambience. Descript has no useful opinion about anything that is not speech, and you will end up in a real DAW anyway. Also shows delivering to a network with a technical spec. The round trip out is the weakest part of the product.
3. Auphonic: best for sounding like the same show every week
Free tier of 2 hours of processed audio a month, with an Auphonic jingle attached to the output. Paid recurring credits are sold by processing hours per month: S 9 hours, M 21, L 45, XL 100, XXL 250. One-time credit packs are also available and never expire. Billing is on the duration of audio you process, with a 3-minute minimum per production. Verified 15 August 2026.
We have quoted hours rather than a dollar figure here deliberately: Auphonic’s own page sells the tiers by processing hours, and the currency figures reported around the web disagree with each other. Check the tier you want on the page before you buy. The hours are the part that decides which tier you need, and the hours are what the page states.
What it does is unglamorous and is the difference between a hobby and a show. Loudness normalisation to broadcast targets, levelling between a host who leans in and a guest who does not, noise and reverb reduction, and consistent output across episodes. Listeners never notice this working. They notice it absent. As the episode they had to turn up and then down again, which is the most common reason someone stops listening to an otherwise good show in a car.
The one-time credits are the genuinely useful pricing detail and are why this ranks above the alternatives. Podcasts are not published on a smooth schedule; a seasonal show or an irregular one gets punished by monthly allowances that reset unused. Credits that never expire match how the work actually arrives.
Who it’s wrong for: anyone already finishing in a DAW with a mastering chain they trust. This is automation of a job you have already solved. And anyone who wants to hear each decision as it is made; Auphonic is deliberately a black box, which is the point for most people and intolerable for a few.
4. Castmagic: best for the work that happens after the edit
Hobby $19/month ($239/year) for 30 upload hours. Starter $48/month ($579/year) for 100 hours. Team $139/month ($1,669/year) for 400 hours. Verified 15 August 2026.
The unglamorous truth about independent podcasting is that shows rarely die from bad episodes. They die from the fortnight where the episode was finished on Sunday and the show notes, the timestamps, the description and the three clips were still not written by Thursday.
Castmagic takes the finished audio and produces that whole layer: notes, chapters, titles, summaries, social copy, newsletter text. The output is somewhere between a first draft and a usable asset, you will rewrite the titles, because titles are judgment, but rewriting is a different task from starting.
Price it against the alternative honestly. The Hobby tier is roughly what a freelancer charges for one set of show notes, for thirty hours of uploads a month. If that layer is currently not happening at all, this is the cheapest entry on this list relative to what it changes.
Who it’s wrong for: shows where the written layer is itself the product. Anything with a substantial newsletter, or a research-led show whose notes carry citations and links that have to be right. Generated notes are confident about details, and confident-and-wrong in a show note is worse than absent.
5. ElevenLabs: best for pickups, ads, and other languages
Free 10,000 credits a month. Starter $6/month for 30,000 credits. Creator $22/month for 121,000. Pro $99/month for 600,000. Verified 15 August 2026.
Three real jobs here, none of them “replace the host”. Fixing a line you got wrong without booking the studio again. Producing localised versions of an episode, which is the only cheap audience-expansion lever in podcasting. And recording sponsor reads in a cloned voice when the ad copy arrives after you have finished the edit, which is the situation that generates most late-night re-records.
Read the licence before you use it on a sponsored episode. Verified 15 August 2026: the free tier does not include a commercial licence; commercial use starts at the Starter plan, $6 a month. Six dollars is not a business decision. But discovering the distinction after a sponsored episode has shipped is, and it is a distinction most people do not know exists. This is the same trap Recraft’s free tier sets for designers.
Disclosure is the other half. A cloned host voice in an ad read sits in territory where audience trust is the entire asset, and some sponsors now ask directly. Decide your policy before a sponsor decides it for you.
Who it’s wrong for: anyone hoping to generate whole episodes. Synthetic conversation is detectable at length and it burns the only thing a podcast has, which is the sense that a person is actually there. Also anyone whose show’s value is reportage or interview. The voice is not the product, the access is.
Shows this list won’t suit
Video-first podcasters. If YouTube is your primary distribution, your stack is a video stack and the ranking changes completely.
Narrative and documentary producers. Your work is sound design, and four of these five treat audio as speech to be cleaned rather than material to be composed.
Network and branded-podcast teams. You have a delivery spec, a legal review, and probably a producer. Your constraint is approval, not production.
Anyone hoping AI makes the show worth listening to. Every tool here removes friction around the conversation. None of them improves the conversation, and the conversation is the entire product. Shows that are struggling are almost never struggling because of their audio chain.
Left off the five
Adobe Podcast Enhance. The best free thing in this category, and genuinely remarkable on bad guest audio. Left out because it is one feature rather than a workflow, and both Descript and Auphonic cover the same ground inside something you are already using. Use it anyway when a guest sends you a disaster, it costs nothing.
Cleanvoice. Focused, good at filler and stutter removal, and squarely inside what Descript already does for anyone editing there. A strong pick if you edit somewhere else.
OpusClip and the clipping tools. Real value, wrong list. Clipping is a video job and it is covered in our list for video editors. If your show is audio-first, clips are a distribution decision rather than a production one.
Podcastle and the all-in-ones. They do a bit of every stage, and the recording stage is the one you cannot afford to have done adequately. We would rather see the recording handled properly and the rest assembled.
Hosting platforms’ built-in AI features. Buzzsprout, Captivate and the rest now generate notes and transcripts. If yours does it well enough, you do not need Castmagic. Check what you are already paying for before adding a subscription.
Podcaster questions
What is the minimum viable stack for a new podcast?
Riverside’s free tier or a local recording, Descript’s Hobbyist plan at $16 a month billed annually, and Auphonic’s free 2 hours. That covers recording, editing and mastering for one subscription. Add Castmagic when the show notes are reliably not getting written, and ElevenLabs only when there is a sponsor or a second language in play.
Can I use AI voice cloning in a sponsored episode?
Only on a paid tier. ElevenLabs’ free plan does not grant a commercial licence; commercial rights begin at Starter, $6 a month, verified 15 August 2026. Separately from licensing, check your sponsor’s position on synthetic reads and your own disclosure policy. The legal question and the trust question have different answers.
Is Riverside worth it if I only interview people occasionally?
If the interviews matter, yes, and it is the last thing to cut. Local recording is insurance against the one failure mode you cannot repair, and the guest who sounds like a drive-through is unrecoverable. If you record alone, skip it entirely.
Why quote Auphonic’s hours instead of its price?
Because its pricing page sells the tiers by processing hours, and the dollar figures reported by third-party sites disagree with one another. The hours are what decide which tier you need, they are stated on the vendor’s page, and a number we could not verify there is worse than no number.
Will AI-generated show notes hurt my SEO?
Not by being generated. They hurt when they are wrong, and generated notes are most confidently wrong about exactly the details people search for, names, figures, product references. Check those and publish; do not check them and you have automated the production of errors under your own name.
Do I need Descript if my host already transcribes episodes?
A transcript is not the same product. Your host gives you a transcript to publish; Descript gives you a transcript you edit the audio with. If you only need the artefact, use what you have.
Last reviewed: 15 August 2026. Every price above was read off the vendor’s own pricing page on that date, with Auphonic quoted in processing hours for the reason given in its section. We re-review quarterly.
Leave a Reply