Last updated: October 2026
Every creator who interviews guests online has one recording they were proud to land and quietly embarrassed to publish. The guest was sharp, the answers were gold, and the audio sounds like it was captured inside a tin can at the bottom of a pool. The guest’s microphone was fine. The problem is what got recorded: not their voice, but a video call’s compressed opinion of their voice.
The short version
- Record at each end, not over the internet. The method is called a double-ender: every speaker is captured locally, on the device in front of them, and the call is only how you talk.
- Use a recording platform, or run it manually. Riverside, Zencastr, SquadCast and StreamYard automate the double-ender in a browser. A guest with a phone voice recorder can do it by hand.
- Send a prep note the day before. Headphones, the softest room they have, the best mic they own, a wired connection if one exists, and a phone recording as backup.
- Run a backup and download the same day. Keep one recorder that answers to nobody’s wifi, and never let a guest close the tab mid upload.
On this page
- Why a recorded call sounds like a call
- The double-ender: record at both ends of the line
- Recording platforms, compared by what they capture
- When your guest can only manage a meeting app
- The prep note to send your guest
- The first ten minutes: checks before questions
- When the connection falls over mid interview
- Backups that answer to nobody’s wifi
- After the call: separate tracks become one conversation
- Frequently asked questions
- Keep exploring
Why a recorded call sounds like a call
Meeting apps are built for conversation, not capture. To keep a call feeling live, they compress each voice hard, strip out lows and background detail, pump the volume up and down, and lower the quality whenever the connection dips. Pressing record on a call saves a faithful recording of that squeezed stream, dropouts included. The microphone was never the bottleneck.
The processing is not a flaw; it is the design. A meeting tool’s job is speech that stays understandable in real time on bad wifi, so it discards everything a conversation does not strictly need. Your ear forgives all of it live, helped by lip movement and context. A viewer meeting the recording later hears exactly what was kept and what was discarded: the warble, the robotic edges, the word that vanished when the bitrate fell.
There is a second problem: most call recordings are a single mixed track. Your voice and your guest’s voice arrive baked together, so a cough under their best sentence, a level mismatch or a doorbell on their side is welded into your side too. Nearly every fix in the edit assumes the voices live apart. On a call recording, they do not.
The double-ender: record at both ends of the line
A double-ender records the same conversation twice, once at each end, on the devices sitting in front of each speaker. You still talk over a call, but the call is only the conversation. Afterwards you download both local tracks, line them up, and throw the call away. Because nothing in the recording travelled over the internet, nothing in it can carry the internet’s damage.
The name is old broadcast craft: radio producers had studio guests recorded locally at both ends of a phone line for decades before the software existed. The principle has not changed. What changed is the labour. Run manually, a double-ender means your guest records on their own device, a phone voice recorder, a portable recorder, or a simple recording app on their computer, while you talk over any call you like. You clap once together, or count in and clap on one, so the tracks have a shared marker to align on later, and the guest sends you their file before you hang up.
Run through a platform, the same thing happens without asking the guest to operate anything: they click a link, the platform records on their device, and the file uploads as you go. Manual costs nothing and works with any guest who can press one button; the platform costs money beyond any free tier and removes the guest’s jobs almost entirely. With two guests recording their own ends as well, you have built a triple-ender, and the arithmetic is the same.
Recording platforms, compared by what they capture
Riverside, Zencastr, SquadCast and StreamYard all run the double-ender for you: guests join from a browser link with no install, every device records locally on separate tracks, and the files upload during the session. They differ in the shape of the tracks and what surrounds them, so choose by your edit and your guests, not by a ranking.
| Platform | How it records | Separate tracks | What you get |
|---|---|---|---|
| Riverside | Locally on each participant’s device, uploading to the cloud in real time as the session runs | Yes, separate audio and video tracks for each participant | Uncompressed 48 kHz WAV audio and video up to 4K, per Riverside’s own help pages |
| Zencastr | Locally on each device, with a cloud backup | Yes, a track for each guest is its headline promise | Video up to 4K and 48 kHz audio, per its homepage |
| SquadCast | Locally at the source, with Progressive Uploads auto-saving the recording as it happens and a cloud backup behind it | Yes, separate locally recorded audio and video files | Raw, uncompressed audio files, per its own site; now part of the Descript family |
| StreamYard | Local recordings on each participant’s device, plus a cloud recording of the session | Yes, individual audio and video per participant locally; individual cloud audio tracks are a separate feature | Local audio exports as WAV; built first for live shows, with recording as a studio mode |
Read the table the way an editor would. If your interviews become video podcasts and vertical clips, the local video ceiling matters and Riverside and Zencastr both document 4K. If the show is audio first, any of the four hands you clean separate WAV level material. If you also stream live with guests in front of an audience, StreamYard’s streaming roots show. SquadCast’s pitch is the upload discipline: the recording saves itself continuously, so a dead laptop at minute forty loses almost nothing.
Two cautions apply to the whole category. Ceilings and track features move between plans and change often, so confirm the numbers on the platform’s own pages on the day you commit, not from last year’s comparison video. And local recording only captures what the guest’s room and microphone give it. The platform removes the internet from the recording; it cannot remove a tiled kitchen.
When your guest can only manage a meeting app
Sometimes the guest knows exactly one tool, the slot is short, and the tool is Zoom. Use it, but set it up like a recorder: turn on the setting that saves a separate audio file for each participant, use its original sound option so the app stops polishing the voice, and treat the result as a backup while the guest’s phone captures the real track.
Both settings live in the desktop app. Under recording settings, “Record a separate audio file for each participant” splits the call audio into one file per speaker, which at least restores your editing moves. Under audio settings, original sound backs off the noise suppression and automatic gain that make voices sound processed, at the cost of hearing more of the room. Neither changes the fundamental fact: the audio still travelled over the internet to reach you, compressed for conversation on the way.
So the honest ranking for a meeting app session is this. The call recording is your safety net and your sync reference. The recording that becomes the episode is the guest’s local one, even if that is only a phone voice memo recorded next to their laptop. That pairing, a familiar app for talking and a local capture for keeping, rescues most low-tech guest situations without a tutorial over email.
The prep note to send your guest
Most remote interview quality is decided before anyone presses record, by one short note sent the day before: where to sit, what to put on their ears, what to plug in, and what to run as a backup. Write it once, keep it warm, and paste it into every booking. Guests are doing you a favour; the note should feel like you are looking after them, not setting homework.
One boundary before the note itself: finding the guest, pitching them and agreeing what the conversation is for is a different job from recording it. Our guide to how to collaborate with other creators covers choosing partners and making the approach. This article starts the moment the guest has said yes, and the note below is the first thing you send.
Here is the note, ready to adapt:
- Join from a computer if you can, with the power cable plugged in, and I will send one link. Nothing to install.
- Wear headphones or earbuds, wired if you have them. This keeps my voice out of your microphone, which is what stops echo.
- Use the best microphone you have. A USB mic is great; earbuds honestly beat a laptop microphone across the room. Sit about a hand’s width from it.
- Pick your smallest, softest room. Curtains, carpet, cushions and soft furniture calm the echo that makes voices sound distant. A room full of clothes is better than a big empty one.
- Plug into the router if a cable reaches; if not, sit near it, and ask the house to pause big downloads and streams for the hour.
- Close anything that dings. Silence your phone, then set it beside you recording a voice memo. That is our backup, and it has saved more than one episode.
- Bring water, and join ten minutes early. We will check sound together, then talk. Nothing to prepare beyond that.
Every line is there because it kills a known failure: the wrong microphone, the echoing room, the notification chime baked into an answer, the wifi fight with someone else’s film stream. Keep it to one screen. A guest who feels briefed relaxes, and a relaxed guest is an audio upgrade no plugin sells.
The first ten minutes: checks before questions
Join early and spend the first minutes on the unglamorous checks while you make small talk: the correct microphone is the one selected, levels survive a laugh, you cannot hear yourself echoing, the backup is running, and a clap lands on every track. Each fault caught here is a fault the edit never meets.
The classic failure is the phantom microphone. Your guest plugs in a good USB mic, joins the call, and the platform quietly uses the laptop’s built-in microphone instead, because it was the default. Do not trust the gear you can see on camera: ask the guest to read out what the settings panel says is selected, and have them tap the mic while you watch the level meter. A scratch test settles it in five seconds.
Then levels and echo. Ask for a loud laugh or their most emphatic sentence; the voice you must survive is their excited one, not their careful test voice. If you can hear your own voice returning, their speakers are feeding their microphone, and the cure is the headphones from the prep note. The voice craft underneath all of this, distance, gain and room, is the same at both ends of the line, and our guide to how to record better vlog audio covers it in full. For video guests, thirty seconds on framing closes the checks: eyes in the top third of the frame, a window or lamp in front of the face rather than behind it. Finish by recording twenty seconds of chat, pausing, and listening back together before the first real question. It feels slow. It is the fastest ten minutes in the whole production.
When the connection falls over mid interview
On a double-ender setup, a dropped connection interrupts the conversation, not the recording. Each device keeps capturing locally and the upload resumes when the link returns. The plan is agreed in advance: recorders keep running, nobody closes anything, you rejoin, and you pick up from the last clean sentence.
Watch what each layer does for you. The platforms upload while you record, Riverside in real time and SquadCast in small auto-saved pieces, so everything up to the dropout is already safe. Riverside’s help pages note that the local files survive even a browser crash. If the guest’s wifi dies entirely, they rejoin from where they are, or you finish the conversation by phone, on a separate phone where the guest’s phone is the backup recorder, while both local recorders keep rolling; the phone is now only carrying the chat, which is the one job it is good at.
The one true killer is a closed tab. A guest who thinks the session has broken may shut the browser, taking an unfinished upload with it. That is why the failure plan lives in the prep note and gets repeated out loud at the start: whatever happens, stay in the session and let me end it. If a track is somehow lost anyway, the call recording or cloud backup fills the hole at lower quality, which is precisely why you recorded it.
Backups that answer to nobody’s wifi
A solid remote setup has three layers: the platform’s local tracks as the main recording, its cloud or call recording as the safety net, and one simple recorder that depends on no connection at all, a phone voice memo at the guest’s end and a second capture of your own microphone at yours.
Your end is easy to forget because the platform feels sufficient. Record your own voice separately anyway, into a portable recorder or a simple recording app beside the session. Host tracks are lost too, usually when the session is closed before the upload finishes, and your questions are half the episode. The guest’s phone memo is the mirror image: started before the call, left alone, stopped after. Neither backup needs the internet, an account or a setting, which is exactly their virtue at the moment everything else is misbehaving.
Then the download discipline. When you stop the session, keep the guest chatting for the minute the upload needs, and download every track the same day rather than trusting the dashboard to keep them. File the tracks with the rest of the project files; our guide to how to back up video footage covers the folder system that keeps them findable in six months.
After the call: separate tracks become one conversation
Download every track, put host and guest on their own timeline tracks, line them up on the clap if they need it, and level each voice on its own before cutting a word. Separate tracks turn the interview edit from rescue into routine: coughs, dog barks and doorbells all leave without touching the other voice.
The working order that suits vlog interviews: align first, while the marker clap or the platform’s own timing gives you a reference; then gain, bringing the quieter guest up rather than dragging everything down; then the cut, tightening answers and removing crosstalk one track at a time. Long manual recordings can drift apart slowly, because two devices keep time at slightly different rates. Our guide to how to sync external audio with video goes deep on drift, claps and waveform sync, and it is the right next read if your double-enders are the manual kind. Once the edit locks, add captions so the conversation is readable as well as listenable; our guide to captions and subtitles for vlogs covers that workflow.
Two neighbours of this workflow are worth naming, because remote interviews grow into both. Going live with the same guest is a different discipline, one take with an audience, and our livestreaming guide covers it end to end. And the same guest filmed in your own studio, two cameras and one room, is a different craft again: our multi-camera vlogging and interviews guide covers that setup, while our guide to how to start a video podcast covers turning a run of interviews into a series of its own.
Frequently asked questions
The questions creators ask before their first remote interview, answered straight.
What is a double-ender recording?
It is a remote interview recorded locally at both ends at the same time. Each speaker’s device captures their own voice, and their picture on a video platform, so the recording never travels over the internet. Afterwards the tracks are lined up together, and the call you talked over is thrown away.
Why does my Zoom interview sound bad when my guest has a good microphone?
Because Zoom records the call, not the microphone. The voice is compressed, processed and adjusted for the connection on its way to you, and that stream is what gets saved. A better microphone feeds a nicer signal into the same squeeze. Record the guest locally and the microphone finally gets heard.
Does my guest need to install software or create an account?
On the dedicated recording platforms, no. Riverside states it plainly in its help pages: guests click the link you send, with no sign-up or installation, and the same browser based joining is how Zencastr, SquadCast and StreamYard bring guests in. A current browser is the whole requirement.
What should I send my guest before a remote interview?
One short note, the day before: join from a computer and plug in power, wear headphones, use the best microphone they have and sit close to it, choose a small soft furnished room, use wired internet if possible, silence notifications, and run a phone voice memo as a backup. The full note is written out in the prep section above, ready to copy.
Can I record a remote interview with just phones?
Yes, as a manual double-ender. Talk over any call, and have each person record their own side on their phone’s voice recorder at the same time, with a clap at the start to sync on. The guest then sends you their file. It asks more of the guest than a platform link does, and the sound is honestly good.
What happens if the internet drops in the middle of the recording?
The conversation pauses; the recording does not. Local tracks keep capturing on each device and upload when the connection returns, and platforms that upload as they go have already saved everything up to the dropout. Rejoin, keep every recorder running, and restart the answer from the last clean sentence.
How do I line the separate tracks up in the edit?
Platform tracks arrive already timed against each other. Manual double-enders use the marker you made at the start: one clap, or a counted “three, two, one” with the clap on one, visible as a spike in both waveforms. Slide one track until the spikes meet, check the end of the interview for drift, and edit from there.
Is WAV worth it for interviews, or is a compressed file fine?
For the recording you keep, WAV. It is uncompressed, so noise reduction, leveling and editing all work on the full signal rather than on what a codec kept. Compressed copies are fine for sending quick previews around. Record in WAV, edit in WAV, and compress only the finished episode.
How many guests can I record at once?
More than a vlog interview usually needs. Riverside’s help pages describe up to ten recorded participants on separate tracks, and the other platforms work in the same multi guest range, though the exact numbers live in plan details that change. For one host and one or two guests, every platform here has room to spare.
Can I record remote interviews for free?
You can start for free two ways. The manual double-ender costs nothing at all: phones, a call, and a file transfer. The platforms offer free tiers or trials whose recording hours and quality ceilings change regularly, so check the current terms on the platform’s own site before planning a series around them.
Keep exploring
The guides that sit on either side of a remote interview: voice craft, going live, captions, and protecting the files.