Two people sat in a room for 94 minutes. That is the raw material. Everything a viewer decides about that conversation happens in about one second, in a grid of competing rectangles, before a single word is heard.
Interview episodes are the hardest format I package. A tutorial has a promise baked into it. A vlog has a place and a face and a story. An interview has two faces, a lot of talking, and a guest name that means everything to 4% of the audience and nothing to the other 96%. Get the packaging wrong and a genuinely great episode dies at 2% click through rate while a worse episode from a smaller channel does 9%.
So here is how I actually approach packaging for interview episodes, including the parts that annoy clients.
Why do interview thumbnails underperform so badly?
Because most of them show the interview instead of the reason to watch it. The default template is two headshots facing each other, a mic in frame, the guest's name in big letters, maybe a logo. That image communicates one thing: this is a podcast. It does not communicate why this specific hour is worth your evening.
There are three failures stacked on top of each other.
Failure one, the guest name does all the work. Names only sell when the name is already a decision. If your guest has a real public reputation, the name is the hook. If your guest is a very smart operator nobody outside the industry has heard of, the name is decorative text stealing 30% of your canvas.
Failure two, two faces means half the size. Every face you add halves the pixel budget for expression. At mobile size, roughly 320 pixels wide in the feed, two shrunken faces read as texture, not as people. One face at full size reads as a person with an opinion.
Failure three, the title describes the conversation. "Episode 44 with a guest on scaling operations" is a filing label. It is not a promise. Titles for interviews have to carry the payoff, because the image usually cannot carry it alone.
That is the argument for treating this as YouTube packaging rather than thumbnail design. The thumbnail, the title, and the first fifteen seconds of the episode are one unit. Separate them and they contradict each other, and the viewer feels it even if they cannot name it.
What actually goes in an interview thumbnail?
One face, one idea, and one piece of proof. That is the pattern that survives testing more often than anything else I have run.
Here is the decision tree I use before I open the file:
- Is the guest a draw? If yes, guest face large, host out of frame entirely, name in a small clean lockup. If no, skip to step two and stop putting the name anywhere.
- Is there a number or a claim? A number the guest said out loud is the strongest asset an interview gives you. "Built to 40 staff, then cut it to 6." Numbers survive shrinking better than sentences.
- Is there a visual artefact? A screenshot, a product, a chart, a physical object from the story. Interviews are visually poor by nature, so one real object does a lot of lifting.
- What emotion is honest? Not shock face by reflex. If the episode is a calm technical breakdown, a shock face writes a cheque the video cannot cash, and your average view duration pays the fee.
Then the composition. Face on one side, claim on the other, high contrast between them, and nothing in the bottom right corner where the timestamp sits. I keep the whole design readable at 20% zoom. If I cannot read it at 20%, it does not ship, no matter how good it looks at full size.
The best interview thumbnail I have ever made had no guest name, no microphone, no logo, and no second person. It had one face and four words. The episode itself was two hours long.
Should the host be in the thumbnail at all?
Usually no, and this is the fight I have most often. Putting the host in every thumbnail feels like branding. It reads as symmetry, and symmetry is quiet.
Two exceptions. First, if the host's face is the channel's actual brand asset, meaning viewers subscribed for the host and recognise them at a glance, then the host earns the space. Second, if the tension of the episode is between the two people, a disagreement, a challenge, a debate, then two faces is the story and you should shoot the composition for confrontation rather than for a friendly split screen.
Otherwise the host in frame costs you face size and buys you almost nothing. Your audience already knows who runs the channel. They clicked because of what is being said, not who is asking.
The same reasoning applies to microphones, headphones, studio furniture, and the desk. They are the set. They are not the story. A stray mic arm across a face has killed more click through than any font choice ever will.
How do you package an episode when the guest is unknown?
You sell the idea, not the person, and you let the face be the emotional carrier rather than the credential. This is where interview packaging gets genuinely interesting, because you have to find the one line in 94 minutes that a stranger would repeat at dinner.
My process is boring on purpose. I read the transcript, or I skim it with timestamps, and I pull every sentence that meets one of three tests: it contains a number, it contradicts common advice, or it is a story with a consequence. That usually gives me 8 to 15 candidates. Then I cut anything that needs context to make sense, because a thumbnail has no room for setup.
What is left is your packaging pool. From it you write the title first, then design toward the title. Doing it the other way round produces pretty images that promise nothing.
Titles for unknown guests tend to work in these shapes:
- The claim with a stake: "He fired his whole sales team and revenue went up"
- The specific counter position: "Why cold email still beats paid ads"
- The named credential without the name: "The engineer who rebuilt a feed used by millions"
- The confession: "I lost $400,000 learning this"
None of them require you to know the guest. They require you to want the answer. If you build the same instinct into your other formats, the logic carries across, which is why the way we handle packaging for vlog uploads shares the same first move: title first, image second.
What about the first frame and the first fifteen seconds?
They are part of the package, and skipping them is why a good thumbnail sometimes produces bad retention. The thumbnail makes the promise. The opening seconds either confirm it in the viewer's first breath or teach them not to trust your thumbnails again.
For interviews the practical version is simple. Whatever line you put on the thumbnail should be audible in the first thirty seconds of the episode, ideally in the guest's own voice. Cold open with the claim. Then your intro. Then the conversation. This costs one editing decision and protects your click through rate from the slow decay that comes from a channel training its audience to expect a bait.
Click through and watch time are read together rather than in isolation, and the creator resources YouTube publishes are worth reading directly rather than through the third hand version circulating on social. If you also publish an audio version, the packaging for the video needs to stand alone, because the platforms behave nothing alike.
Do interview episodes need thumbnail variants?
Yes, more than any other format, because the guess rate is higher. With a tutorial I can usually predict the winner. With an interview I am choosing between "sell the guest" and "sell the claim", and those are different videos to the algorithm.
We run this as an add on, an A/B variant at plus $15 on top of the standard price, which is $250 for 8 custom thumbnails, about $31 each, one per video, with unlimited revisions and a dedicated designer. Title consultation is included, because for interviews the title is doing half the work and pricing it separately would be dishonest. Turnaround is 48 to 72 hours per thumbnail, and there is a rush option at plus $20 if a guest drop is time sensitive. A title pack, several angles to test against the same design, is plus $25.
Every design is built for the specific episode. We do not run templates, and we test styles on our own channels before they go anywhere near a client's channel. If you want to judge that quickly, our work is the fastest way.
One warning on variants. Do not swap the thumbnail on a video that is currently performing well. Swap on the flat ones, and give each version a real window before you judge it. When you do run these tests, write them down properly, which is a whole discipline in itself and one I covered in how to report packaging results to a client.
What do you do with an interview back catalogue?
Repackage it, because for most interview channels the archive is the cheapest growth available. You already paid for the studio time, the guest's calendar, the editing, and the upload. The episodes are sitting there with a 2% click through rate and a title that says "Episode 17".
Interviews age unusually well. A conversation about a business decision from three years ago is often still true. What has aged is your packaging standard, which was probably worse then than it is today. That gap is free performance.
I would start with the twenty episodes that have the best average view duration and the worst click through. High retention plus low clicks is the exact signature of a good video with bad packaging. Repackage those first, before you touch anything else. The mechanics of doing this at scale are the same ones behind repackaging tutorials and evergreen guides, and there is a timing layer on top of it that I set out in seasonal repackaging of old videos.
While you are in there, be honest about style drift. If your thumbnails from two years ago look nothing like your thumbnails now, the channel page reads as three different shows. Consistency is not a design preference at that point, it is a recognition problem. What styles are actually working right now is a moving target, and I keep notes on it in our writeup on youtube thumbnail trends.
Is a template pack good enough for a weekly interview show?
For a while, yes, and then it quietly stops working. A template gives you a consistent frame and fast turnaround, which matters when you ship every Tuesday. What it cannot do is respond to the episode.
The problem is that interviews vary enormously in what they offer. One week you have a well known guest and a boring conversation. The next week you have an unknown guest who says something remarkable. A template treats both identically, and treats both as "podcast episode with two headshots". That is exactly the image the viewer has learned to scroll past.
There is a longer comparison of the tradeoffs, including where templates genuinely make sense, in our piece on stock templates vs original thumbnails. Short version: templates are fine for shorts and clip channels, and they are a slow leak on a flagship interview show.
We work from Dubai, async, with channels worldwide. If you want a second opinion on your last ten episodes before you commit to anything, a short email to samy@pixelsamy.com is the whole process, and if your bottleneck is the edit rather than the packaging, our sister service also edits video podcasts.