How to Edit a Podcast: Cut the Transcript Before the Timeline

Somewhere in a shared drive right now sits a 94 minute conversation that needs to become a 38 minute episode by Thursday. We've all stared at that file. The editor hasn't opened the DAW yet. They've opened a transcript instead. The fastest way to edit a podcast is to read the conversation before you touch the audio: mark the strongest lines in the transcript, cut the rest on the page, and only then build the timeline from what's left.

Read First

Why read a podcast before you ever open a DAW?

Because reading is faster than listening, and a transcript turns ninety minutes of talking into something you can scan in five. A raw conversation hides its problems in real time: the tangent that goes nowhere sounds fine in the moment, and the fifteen minutes where energy sags only becomes obvious once you can see the shape of the whole thing on a page. Mark the transcript first, pull the strongest lines, and the scrubbing nobody wants to do on a Friday afternoon mostly disappears. This is the same logic behind a paper edit on a documentary shoot: decide the structure where it's cheap to change your mind, then build the timeline from a decision that's already been made.

What To Cut

What should actually come out of a raw conversation?

Five things, almost every time. Dead air and pauses longer than two or three seconds rarely survive a listener's patience, so trim them. Repeated stories, where a guest tells the same anecdote twice because the recording ran long or the thread got lost, only need one pass to make the final cut. False starts and rephrased sentences clutter a transcript in a way that's genuinely hard to catch by ear but obvious on the page, which is half the argument for reading first. Tangents that never reconnect to the episode's actual subject can usually go entirely, not just get trimmed. And filler words, the ums, the you-knows, the likes, get thinned rather than erased: a host with every single filler word surgically removed starts to sound like a hostage read instead of a person having a conversation. (For a deeper pass at that last one specifically, there's a full breakdown in how to remove filler words without making a guest sound scripted.)

Real-World Discipline

How do professional shows decide what survives the cut?

This American Life is the reference point most audio editors reach for, and for good reason: the show is known for restructuring interviews around the strongest narrative thread rather than the order things happened to get said in the room. Host Ira Glass has discussed with Westword how the show decides, episode by episode, which stories survive and which get killed well after the tape stops rolling, a decision that's never really settled at the recording session itself. The lesson for a weekly conversational podcast isn't to rebuild every episode as a narrative documentary. It's that professional editors treat a recording as raw material to be shaped, not a finished product to be lightly tidied. A transcript is what makes that kind of restructuring possible without re-listening to the whole thing three times to find where a sentence actually lives.

Audio Or Video

Does a video podcast change what you cut?

It adds a second layer of decisions on top of the first one. An audio-only episode is edited almost entirely on pacing and content: what stays, what goes, how it flows. A video podcast adds camera cuts: when to hold on the person talking, when to cut to a reaction, when to switch angles on a laugh that lands. If the show records multiple camera angles, do the transcript-based content cuts first and layer the camera switching in the NLE once that spine is locked. Even all-in-one recording tools built for this format lean the same direction: a review of Riverside's plans notes that its paid tiers all ship with a text-based editor and transcription built in, not bolted on after the fact, which says something about where the industry has settled on where editing decisions should start. Doing both passes at once means re-deciding the same cut twice, once for words and once for coverage, and that's where a Tuesday afternoon edit turns into a Wednesday morning one.

A Worked Example

What does this actually look like on a real episode?

Picture a 75 minute interview episode with one guest, recorded on a Tuesday for a Thursday release. The transcript lands within the hour. A first read takes about fifteen minutes, during which the opening six minutes of small talk gets marked for the bin, a story about a failed product launch gets flagged as the episode's spine, a tangent about a conference in Austin gets marked to cut entirely because it never reconnects to anything, and a strong closing thought that got buried at minute 58 gets flagged to move toward the end where it belongs. None of that required touching a waveform. The selects then get strung together as a stringout, played back once in order to check that the jumps actually make sense out loud, and only then does the audio polish (noise reduction, levels, music beds) happen in the DAW. The part that used to take two hours of scrubbing now takes fifteen minutes of reading and one listen-through. That's the trade a transcript makes possible.

From Selects To Social

What happens to the parts that don't make the final episode?

They don't have to be wasted. Every episode throws off three or four moments that are too short or too tangential for the main cut but genuinely good on their own: the aside that got a real laugh, the answer that's quotable out of context, the two minutes where a guest says something sharper than anything in the rest of the hour. Pulling those by hand means a second pass through the same transcript, reframing each clip, writing a caption, and figuring out what to say about it on each platform, and we'd rather spend that time on the next episode. ScriptCut Social takes the long recording, or a cut already built in ScriptCut, finds the moments worth posting, frames them on the speaker, writes a caption and a platform-specific description for each one, and schedules the posts, with one link per client to approve before anything goes out. It posts directly to LinkedIn today; everywhere else, it prepares the clip and copy and you push it live. For a deeper look at this exact handoff, see how to turn a podcast into clips and reels and how to batch edit a full podcast season when it's not one episode but twelve.

Workflow That Scales

How do you edit episode forty the same way you edited episode four?

A workflow that works once and collapses by episode twenty was never really a workflow. It was luck. The version that scales looks identical every single week: transcribe on ingest, mark cuts in the transcript while the conversation is still fresh in someone's head, export the selects into whatever tool builds the final file, then run one listen-through pass for pacing before it ships. That's the entire loop, repeated weekly, and it doesn't depend on one producer's memory of where the good eleven minutes were hiding.

This is the exact middle section ScriptCut is built for. AI Select marks the strongest moments across the raw transcript automatically, AI Arrange puts picks in a sensible order, and Cleanup thins filler words without stripping a host's voice down to nothing. A producer can still override every single pick, play any selected clip back to check the tone actually lands (word-level timecode makes every selection a precise, checkable cut, not a guess), and send the resulting story to a host or client on a share link for approval before anyone opens an editing app. Once it's approved, the arranged timeline goes out to DaVinci Resolve or Premiere Pro through the ScriptCut panel, or out as XML for Final Cut Pro and Avid, so the person doing the final audio polish or camera-switching inherits a decision that's already been made, not a blank 94 minute file. (Resolve itself keeps adding ground here: Blackmagic's own release notes for the current 21.1 update point at more AI assisted organizing tools inside the NLE, on top of whatever arrives already decided from the transcript stage, per Resolve's release history.)

Common Mistakes

Where do podcast edits go wrong?

Over-editing tops the list: stripping every pause and every filler word until a real conversation sounds like two people reading a script at each other. Under-editing is the opposite failure, publishing something close to the raw 90 minute file because trimming felt like too much work that week, and losing listeners somewhere in the first ten minutes as a direct result. The third, quieter mistake is editing live while the notes from the recording are still incomplete, which guarantees re-listening to the same fifteen minutes two or three times instead of once, because nobody marked where the good part actually started.

Where This Fits

Where does this step sit in a normal podcast pipeline?

Most podcast production guides jump straight from "record" to "open your DAW," and skip the part where somebody actually decides what stays. That decision, made on the page before the waveform, is the difference between a show that ships on time every week and one where editing day always, somehow, runs long. It's also the same move documentaries, news packages and corporate video teams make before they ever touch a timeline, which is why terms like radio cut and assembly cut show up across so many kinds of non-fiction editing, not just podcasts.

Need to get captions right once the cut is locked? The free SRT to VTT converter handles the format conversion and cleanup without another signup.

Here's how we'd start a new episode on a tight week: transcript first, selects marked before a single cut is made, and the story approved by whoever needs to approve it before anyone opens an editing app. That's the whole workflow ScriptCut is built around, and it's the fastest way we know to get a weekly show out the door without the editing day running long.

Sources

• This American Life's Ira Glass on interview process
• DaVinci Resolve (Wikipedia)
• Blackmagic Design: DaVinci Resolve 21.1 announcement
• Riverside.fm review and feature breakdown