Last month I opened a project where the host had recorded a double ender for the first time. Two audio files, two video files, one call recording as backup. He was nervous about it. It took me about four minutes to sync the whole thing, and the finished episode sounded like the two of them were sitting in the same room in Chicago. He had been publishing compressed call audio for two years before that.
So let me explain the thing plainly, because most explanations of double enders are written by people who have never had to fix one.
What a double ender actually is
A double ender means each person records themselves, locally, on their own machine, at full quality. Nothing goes through the internet. The call itself, whatever you use for it, is just the conversation channel. Afterwards both people upload their local files and the editor lines them up.
The name comes from radio. Two ends of the line, two recorders, one interview. Nothing modern about it. What is modern is that the tools now do it for you: most remote recording platforms record locally in the browser and upload in the background while you talk.
The payoff is simple. Internet audio is compressed to survive the trip. Local audio is not. The difference between a local track and a call track is not subtle, it is the difference between a show that sounds produced and a show that sounds like a meeting. Every host who switches hears it on the first episode.
Why the call recording is not good enough
I get this argument a lot. The Zoom recording sounds fine, why add steps.
It sounds fine while you are on the call, because your brain is doing the work. Listen back on headphones and you hear it. The codec throws away the top end, so voices go slightly nasal. Noise suppression pumps, so room tone rises and falls between sentences. Worst of all, most platforms duck one speaker when the other talks, which means any overlap, any laugh, any "yeah exactly" gets chopped. You cannot restore that in the edit. The information is gone.
Then there are the dropouts. A call file bakes in every hiccup on either side of the connection. A local file does not know the internet exists. When we take on remote podcast editing work, the first thing I check is whether we have local tracks, because it decides how much of the episode I can actually save.
The call recording is a safety net, not a source. Keep it, never publish it.
The three ways a double ender goes wrong
Almost every double ender problem I have seen falls into one of three buckets.
One: the guest never uploads. The browser tab gets closed before the upload finishes, or the app is quit, or the laptop sleeps. The file exists on their machine, or it existed, and now the guest is a busy person who does not answer emails. This is the most common failure by a wide margin and it has nothing to do with technology.
Two: the tracks drift. Two different machines, two different clocks. Over a ninety minute recording, cheap interfaces and some laptop mics can drift by a few hundred milliseconds. Sync the top of the episode and the end is visibly off. Fixable, but somebody has to notice, and if nobody notices you ship an episode where the guest's mouth stops matching their voice around the sixty minute mark.
Three: the guest records themselves badly. Local recording captures whatever the mic captures. If they are on laptop speakers with the mic picking up their own output, you now have a high quality recording of an echo. Local does not mean good. It means uncompressed.
None of these are reasons to skip double enders. They are reasons to have a checklist, which is why we keep a pre-record checklist and hand it to hosts before their first session with us.
How I actually sync them
The boring method still wins. Everyone claps, or counts, at the top. One sharp transient on both tracks and the alignment is obvious on the waveform in two seconds.
If nobody clapped, I sync to the call recording. The call file has both voices in it, so I use it as a reference bed, drop each local track underneath, nudge until the phasing disappears, then mute the reference. Takes longer. Works every time.
For drift, I check three points: start, middle, end. If the end is off, I stretch the track by the small amount needed rather than cutting and re-syncing chunks, because cutting means every clip after the cut moves and any b-roll you placed goes out of position.
A human doing this catches things software does not. Automatic sync tools line up the loudest transient and call it done. They do not notice that the guest's file starts ninety seconds late because they joined the call, left, and rejoined. We do not run any of this on silence detection. An editor watches the whole episode, which is the only way to catch a drift that shows up at minute sixty two.
Double enders and video
Audio is the reason most people start. Video is where it gets interesting and where it gets annoying.
Good: each person is recorded at full resolution instead of a 720p compressed stream. Faces look like faces. You can push into a shot for emphasis and it holds up.
Annoying: two different cameras, two different rooms, two different white balances, and one of them is sitting under a window at golden hour. Local video makes the mismatch more obvious, not less, because there is enough detail to see how different the two images are. That is a grading problem and it is solvable, but it is real work. I wrote about the approach in matching cameras that do not match, and it is the same logic we use on multi-cam editing for in person shows.
Also: local video files are big. A ninety minute 4K local recording from both sides is a lot of gigabytes moving across the world. Plan the upload before the session, not after. If the guest is in a hotel on conference wifi, you are not getting that file tonight.
When I tell hosts not to bother
Here is my actual opinion, and it is less enthusiastic than most articles on this topic.
If your show is audio only, weekly, and your guests are strangers who booked through a form, a full double ender with video is often more friction than it is worth. You will lose files. You will chase people. You will publish late. A solid platform recording plus one good local track from the host covers most of the gap, because the host is on every episode and the host's voice is the one listeners hear most.
Do the full double ender when:
- The episode is going on YouTube and faces matter.
- The guest is a repeat collaborator or a co-host, so the handoff is habit.
- You are cutting vertical clips from it, where quality gets scrutinized frame by frame.
- The conversation is high value enough that you would re-record it if it failed.
Skip it, or do a half measure, when the guest is a one time booking with no technical patience and the show never leaves the audio feeds. Being honest about that has saved hosts a lot of misery. If you are planning around clips, the recording decisions matter even more, which is the whole argument in shooting for clips while recording.
The part nobody sets up until it hurts
Double enders multiply your files. One episode is now four or six assets living in three places: the guest's cloud, the platform, your drive. That is fine until a drive dies or a free plan expires and the recording goes with it.
Set the backup rule before you record, not after your first scare. I have opinions about this that I put in backing up podcast footage, and the short version is that a file existing in one location is not a file that exists.
The other overlooked piece is turnaround. More files means more time before anyone can start cutting, because the edit cannot begin until the slowest guest uploads. We hold a 48 hour first cut, and the honest constraint is that the clock starts when the last file lands, not when you finish talking. That distinction is most of what makes our 48 hour promise possible at all.
What I would tell a host starting this week
Record locally on both ends. Clap at the top. Keep the call recording as insurance. Make the guest confirm their upload finished before they close the tab, in writing, on the call, while they are still paying attention. Name the files with the date and the person.
That is the whole discipline. It takes ninety seconds per episode and it removes the single most expensive category of podcast disaster, which is a great conversation you cannot publish.
And if the show is heading to YouTube, remember the recording quality only carries so far. The click comes from the packaging, which is why our sister studio ClickTheory exists in the first place.
If you want a second pair of eyes on your remote setup before your next session, write to me at samy@pixelsamy.com. I am in Dubai, I work with hosts everywhere, and I would rather talk you out of a workflow than fix it after the fact.