In case you’ve been living under a podcasting rock, Descript is an all-in-one audio and video editing software. And yeah…I’m an audio person. So I’m gonna stick to the stuff I know.
Like many of you, I have lost DAYS of my life to various Descript dramas.
I have watched in terror as it has glitched, hiccuped, and out-and-out failed.
I have had my workday interrupted by constant updates and unnecessary changes to the UI.
Seriously, do you know how hard it is to write up a list of instructions for a client when buttons and menu items change every few weeks?
And I have shaken my fist at the sky as I have lost work…over and over and over again.
I truly hate this program.
And yet…
About a month ago, I was meeting with a client who has a little bit of grant funding for an interview-based show. She was asking for ideas on how to produce her podcast on a limited budget. And I found myself saying…
Have you thought about Descript?
Last week, when a colleague shared how she had been up all night, pulling together a scratch mix for a narrative episode, I told her….
You should use Descript for that.
And then the other day, I was coaching a client who still transcribes by hand, as a way to get more familiar with the tape. And I was shocked by the inefficiency of it all, even though that was exactly my philosophy for YEARS – before decent AI transcription came around. And so I said…
Seriously, you gotta embrace the 21st Century and get yourself a Descript account!
(Okay, to be honest, I might have suggested Trint as a first option, because all she wants to do is transcribe. But Descript was a close second.)
I’m not exaggerating when I say I have a hate-hate relationship with this program. But…until we have a better option -- or until Descript decides to take the needs of audio makers seriously -- let’s talk about how we can best use this program for our needs.
And just a little aside. There is nothing that I’m going to mention here that I haven’t already said directly to Laura, Descript’s VP of Product. A few months ago, Laura offered to meet with anyone who wanted to share their concerns. And I took her up on that offer. And she was lovely and gracious and I have nothing negative to say about the interaction.
Okay, let’s start with the simple stuff. Descript claims to be an all-in-one solution. You’re supposed to be able to use it to record, transcribe, organize, edit, assemble, process and mix your podcast. All in one place! All with one tool! All for the low-low price of $24/month.
Yes, there is a cheaper option, but let’s stick with the option that they claim is “most popular.”
So, let’s unpack.
Recording
So, yeah, you can record using Descript. Descript owns Squadcast. And Squadcast comes for free with your Descript subscription.
But here’s the thing – and I say this after tons of testing and way too many real-world failures – Squadcast does not sound as good as Riverside.
This is going to be controversial, so I’ll say it again.
Squadcast does not sound as good as Riverside.
I could probably write a whole newsletter about my hate-hate relationship with Riverside. Believe me. I get it.
But the average Riverside recording just sounds better than the average Squadcast recording.
I cannot tell you the number of Squadcast recordings I’ve gotten back that are glitchy. Audio fades in and out. Weird lag. Strange artifacts.
These things are not supposed to happen. The audio is supposed to be recorded locally and uploaded in the background.
But I’m here to warn you…do not trust Squadcast. It will bite you in the ass.
Transcribing
Okay, if you want an accurate transcription of your project, don’t use Descript.
Use Trint. Maybe Temi.
Not Otter. I know it’s free/cheap. But you get what you pay for.
Descript is only so-so when it comes to the accuracy of the transcription. Worse yet, it’s really annoyingly difficult to correct.
And sometimes, when you try to correct your transcription, the program will just refuse to accept the change. Completely. Like the computer can’t admit that a human understands human speech better than it does.
But…
How many of us actually need 100% accurate transcriptions of our projects? I know I don’t! And if you're gonna use some of Descript's other features, you might as well make use of the free transcriptions that come with it.
For me, this is a minor annoyance. Not the source of my hate-hate.
Organizing
What is it that Peter Parker learns from Uncle Ben – or Aunt May, if you’re from the Tom Holland Universe?
With great power comes great responsibility.
So, for me, this is one of Descript’s greatest superpowers.
Descript allows you to organize your audio into “projects.”
This makes copying, pasting and editing audio FROM MULTIPLE INTERVIEWS as simple as moving words around on a text document.
Better yet…you can search for a particular word or phrase across all the compositions in a project.
So, when we were working on Believable and I needed to find every piece of audio that mentioned Coco’s blonde hair and blue eyes, all I had to do was search for the word “blonde” or “blue.”
And if someone mentions an important date, but I can’t remember who, I can simply search for a keyword that would be found in the same sentence.
It’s sooo good.
But here's the problem.
Each project can hold an indeterminate number of compositions.
How many?
Your guess is as good as mine!
I’ve searched online for this information. I even asked Laura, Descript’s VP of Product. And she told me that her developers aren’t sure how many compositions (or how many gigs of audio) a project can hold before it starts getting buggy.
When projects get too big, those really annoying errors I mentioned earlier start cropping up. And there’s no way to make sure you stay within an appropriate project size, because there’s zero guidance on how big is too big.
And yeah, the company will say that the solution is to start a new project for every episode. But if you’re working on a serialized narrative, like I often am, that’s not actually gonna work.
Editing
Okay, look, I’m just gonna say it.
Descript’s edits sound like crap.
Terrible, terrible crap.
If you edit your podcast with Descript and I happen to hear it, chances are I will stop listening halfway through the first episode. The edits are just soooo bad, my ears can’t take it.
And yes. I will JUDGE you for every single one of those bad edits.
It’s a wonderful dream to imagine that you can edit a text document and end up with a beautiful sounding piece of audio. But the technology just isn’t there yet.
And yes, it is possible to finesse edits using the “timeline” feature, but I find that I just can’t get the granularity I need to make things sound really good.
It’s really great for a non-audio person to be able to go into a transcript and make rough edits for content. I recommend this to my clients who want to control their content -- and costs -- without becoming an expert in traditional editing software.
But I always – always – export the audio timeline into another editing software (ProTools or Adobe Audition) where I can clean up the edits and finish the rest of the mixing and mastering process.
And you know what? It works. And it saves me a TON of time. So even though Descript's edits are crap, this is something I’m willing to work with. Because I can fix crappy edits. After all these years in audio, I’m kinda brilliant at it.
But before we move on – I want to warn you about the WORST thing about Descript’s editing tools.
Descript has a whole bunch of super-enticing options in its awkwardly named AI-enhanced “underlord” menu.
Edit for clarity Remove filler words Remove retakes Shorten word gaps
Now, all of these are things that you might want to do to your audio. But I’m gonna say this, and I want you to hear me clearly.
Back away.
Do not touch.
Danger, Will Robinson!
DANGER!
All of these options sound terrible. Like really, really terrible.
Not only does the AI try to make edits that just are never going to sound natural (after all, not every edit is possible) but also, it tries to take a lot of the “humanness” out of your podcast.
People use filler words. They sometimes repeat themselves for emphasis. They pause when they’re overcome with emotion. They ramble when they’re feeling verklempt.
Don’t let the AI underlord remove the humanity from your podcast!
After all, reality is the only thing that's real.
(I'm on a role with these geeky pop culture references. A gold star to the person who finds them all!)
Assembling
Okay, I’m sure by now you’re wondering why I actually recommend this god-forsaken program to so many of my clients.
And it’s this -- right here. This is it!
Descript can save you dozens of hours of work – per episode – when it comes to pulling cuts, rough editing cuts, recording scratch tracks, assembling scratch mixes and assembling the final timeline.
For those of us in the serialized narrative world, Descript is a GAME CHANGER.
As promised, here’s my process.
Step 1: Record in Riverside. We’ve already discussed the reasons for this.
Step 2: Upload separate audio files for each voice to Descript using the “combine into a multitrack sequence” option. That way, you can still edit and process each voice separately later on.
Step 3: Organize your audio into projects. Because it’s not really feasible to organize by episode, I generally organize by type. So for my most recent gig, I had four projects. Host Reflections, Interviews, Interviews with Doctors, and Archive Audio.
Step 4: Highlight by episode. Once I have a pretty decent idea of my episode structure, I go into each interview and highlight the clips I might want to use, color coded by episode. So E1 might be yellow, E2 orange, E3 pink, etc, etc.
Descript, if you’re listening, I would again like to repeat my request for more highlight colors. Many serialized narrative podcasts have eight episodes or more. But last I looked, you only offered six different colors. This seems like an easy thing to fix.
Step 5: Copy selects into episode projects. Once I have highlighted all the interviews for an episode, I use the “copy highlights” function to copy the episode selects for each episode into new projects.
I use a different project for each episode, otherwise the projects will quickly get too big and start acting buggy. Trust me on this.
Step 6: Copy the episode selects into a new composition, and start making choices. Delete stuff you don’t need. Put clips in order. Figure out your final episode structure. I make a copy before digging in, so that I can always go back to the original, if I miss up.
Step 7: Write your narration. Yes, I know that Descript allows you to write scripts inside the program, but have you ever tried to actually do it? The lag will drive you batty. Better to copy everything into Google or Word and do your writing there.
Step 8: GENERATE YOUR SCRATCH TRACKS. Oh Lordy. This is amazing. I do not see myself ever using Descript’s AI voice generator for its intended function – which is to generate the sound of the host’s voice, so that the host does not need to bother to record their own podcast. But for scratch tracks? It’s brilliant.
This alone will save you multiple hours -- per episode.
Step 9: Get feedback. Make changes. Repeat, as necessary. Seriously, you can generate a new scratch track in a matter of minutes, not hours. So be brave! Be bold! Try out that crazy idea you’re not sure will work. What do you have to lose?
Step 10: Record your real host tracks. You can either edit and assemble them in Descript or bring them into your audio software. I’ve done both, and haven’t noticed a huge efficiency either way.
Step 11: Export your timeline to ProTools, Logic or Audition. And finish your mix from there!
Okay, I know that’s a LOT of steps. But count up the number of steps that it takes to do this all manually, and I’m sure you’ll agree…you’ve got some serious time savings here.
Time is money. And in an industry that’s getting squeezed, we need to take advantage of efficiencies where we can.
Mastering
I’m gonna quickly mention this particular feature of the stupidly named AI-powered “underlord” separately, because it is actually useful.
Descript offers something called “Studio Sound” which is very similar to Adobe’s free AI Enhancer.
This will take your audio and run it through a bunch of filters to make it sound more present and less roomy.
Studio sound is definitely worth exploring, but it’s not a one-size-fit-all solution for every piece of bad audio. And most of the time, you want to run it at WAY lower than full strength.
Again, we listen to podcasts to hear humans, not robots. So make sure your voices still sound like people!
Mixing
Okay, I’m gonna admit it. I have ZERO experience mixing in Descript. I always click the “export timeline” button and open up the session in Pro Tools, where I really have the control I need to make it beautiful.
When you first open a Descript-generated session, you’re going to hate your life choices. It’s gonna randomly create a whole bunch of unnecessary audio tracks, and you’re gonna have to spend some time moving things around getting organized.
But quite quickly, you’re gonna find yourself looking at a really decent assembly that you can alter and improve as you please.
The joy of this moment cannot be overestimated. You have literally skipped over the tedious, annoying and soul-draining steps of setting up a mix, leaving you the brain space to focus on the creative and soul-affirming parts.
Seriously, I know we’re all worried about AI taking our jobs. But so far, it’s only taking the parts of my job that I never enjoyed in the first place.
No matter how hard I try, I can’t be mad about that! |