I pulled up an episode last month where the guest's name card appeared eleven times in fifty two minutes. Eleven. Same guest. Same chair. The editor before us had set it as a template and let it fire on every camera change, and the effect was that a stranger kept walking into the room to introduce someone who had been talking for forty minutes.
That is the failure mode nobody warns you about. Lower thirds do not usually fail by being ugly. They fail by being present.
A lower third is a piece of information delivery, not a design opportunity. Once I started treating it that way, the decisions got easy, and our episodes got quieter in a way hosts noticed before they could name it.
The four second rule I actually use
Read the name out loud at a normal pace. Add a beat. That is roughly two seconds for most names and titles. So the card lives on screen for four seconds: a quarter second in, three and a half readable, a quarter second out.
Four seconds sounds short in the timeline. It is not short on screen. When a viewer sees a moving object in their peripheral vision, they look at it. If the card is still there after they have read it, they look again, and now they are auditing your graphic instead of listening to your guest explain how their first company died.
I hold longer in exactly two situations. Long titles that genuinely need a second line, and non Latin scripts where the reader may need more time. Everything else is four.
If your lower third needs a hold longer than the sentence being spoken underneath it, the problem is the copy, not the timing.
Where it goes matters more than what it looks like
Most podcast frames are two people in medium shots, and most default templates were designed for news desks. So the card lands directly across the speaker's chest, or worse, over the mouth of the person in a two shot.
Here is my placement checklist, in the order I run it:
- Bottom safe area, but not hugging the true bottom edge. Leave room for platform chrome. YouTube's progress bar and a phone's caption track both eat the same real estate.
- Left aligned under the person being named. If the guest sits camera right, the card goes camera right. Sounds obvious. Gets ignored constantly.
- Never crossing the vertical center line in a two shot. That center gap is where the eye rests between faces, and putting text there is like standing between two people in conversation.
- Clear of the busiest part of the background. If there is a shelf, a plant, or a neon sign back there, the card fights it.
On a two camera or three camera show this gets more interesting, because the safe zone changes per angle. That is one of the quiet arguments in favour of running three cameras instead of two: the wide gives you a place to put graphics that does not sit on anybody's face. We build the card position per angle, not once per show, and it takes maybe ten extra minutes across a whole episode.
One animation. Not three.
Open almost any template pack and you get a card that slides in, has a bar that wipes, has a subtitle that fades on a delay, and then has a little accent line that draws itself. Four separate motions to deliver eight words.
Pick one. My default is a short vertical rise with a fade, around twelve frames, with a soft ease out. No bounce. No overshoot. Bounce reads as playful in a promo and reads as amateur under a serious answer about layoffs.
The one exception I make is a mask reveal that wipes in the same direction the person is facing, because it feels like the graphic is coming out from behind them rather than landing on top of them. It is a small thing and most viewers never consciously register it, which is exactly the point.
The exit deserves as much thought as the entrance and almost never gets it. A card that snaps off screen is more distracting than one that slides in. I usually exit slightly faster than I enter, on a fade only, because by then the viewer has already stopped looking and I do not want to call them back.
The three moments a name card earns its place
My honest position: most episodes need three lower thirds. Sometimes two.
- The guest, once, roughly fifteen to thirty seconds after they first speak. Not at the cold open, where the viewer is still deciding whether to stay. Once they are in.
- The host, once, if the show is still building an audience. Established shows can drop this entirely. If people already came for you, introducing yourself is odd.
- A credential callback at the moment it becomes relevant. The guest mentions running trials at a hospital, and their title tells you they ran the department. That is a card doing real work.
Everything else is decoration. Chapter markers do a better job of the thing people think repeated name cards do, and they cost you nothing in screen space. If somebody joins forty minutes in from a shared timestamp, they do not need a graphic, they need the episode structured well enough to catch up, which is where good chapters beat graphics every time.
Type is where amateur hour shows up
The fastest way to make a lower third look homemade is to set it in the same weight as everything else on the internet. A few things I hold to:
Name and title need a real hierarchy. Name at full weight, title at maybe seventy percent of the name's size and a lighter weight, or the same weight at sixty percent opacity. If both lines shout, neither reads.
Size for the phone. I check every card at twenty five percent zoom on the timeline, which approximates a phone in a hand. If I cannot read the title at that size, the title is too long or too small. Most of our viewers are watching on a screen smaller than a paperback.
Contrast against the actual footage, not against a mockup. This is where lower thirds and grading collide. A card that reads perfectly over a bright wall vanishes over a dark bookshelf, and if your two angles are graded differently the same card behaves differently on each. It is the same class of problem as matching cameras that do not match, and it is why we place graphics after the grade is locked, never before.
If you want a card to survive any background without a heavy box behind it, add a very soft drop shadow at low opacity and a large blur. Not a hard offset shadow. A shadow you cannot see, only feel.
Colour: borrow from the room, not the brand deck
The brand colour is rarely the right colour. Brand palettes are built for slides and websites with white backgrounds. Your episode is a warm room with skin tones in it, and a saturated brand blue over that reads like a sticker.
What I do instead is sample the footage. Pull a colour from the set, desaturate it, then push it a little further from the skin tones so it separates. You end up with a card that looks like it belongs in the room. The brand mark can still sit in there small, in white, and nobody will complain that you underused the palette.
This is also where colour work on a two person show pays off twice. A consistent grade means one graphic package works across every angle, and you stop building per shot exceptions.
What changes for vertical
Everything. Vertical clips are a different medium and a lower third built for 16:9 does not survive the crop.
On vertical, the bottom third belongs to the caption track and the platform UI. So the name card moves to the upper third, gets bigger, and usually gets shorter. Name only. Nobody reads a job title in a nine second hook.
Half the time I drop the card entirely on clips under thirty seconds and let the captions carry it, because a clip is trying to earn three more seconds of attention and a graphic spends them. When we cut vertical clips from a full episode, the graphics package is rebuilt from scratch rather than scaled down. Scaling down a horizontal card is how you get text that is technically present and functionally invisible.
Same logic applies further up the funnel. The thumbnail and the title carry the packaging job on YouTube, and no lower third has ever saved a click. If that side needs work, our sister studio ClickTheory handles thumbnails and does nothing else.
The audio parallel nobody makes
Here is the thing that made this click for me. A lower third is the visual version of a sound effect.
When you add a whoosh under a graphic, everybody agrees that ten whooshes an hour is too many. Nobody applies the same math to the graphic itself. But attention is one budget, and a card that pulls the eye costs the same as a sound that pulls the ear.
Our editors watch the whole episode before touching a single graphic, which is how we catch things like a card landing on the exact word the host was building to. You cannot catch that by scrubbing. It is the same reason we do not auto cut by silence detection: the machine finds the gap, not the meaning. That principle runs through the audio cleanup work too. Restraint reads as quality, in both channels.
Build one card. Place it per angle. Four seconds. One animation. Three appearances an episode. Colour sampled from the room. Rebuilt, not scaled, for vertical.
The best compliment a lower third gets is silence. Nobody emails to say the name cards were tasteful. They just tell you the episode felt more professional than the last one, and they cannot point at why.
If you want to see what your show looks like with the graphics pulled back, how we work through an episode explains the order we do things in, and the first cut lands in 48 hours. Send me one episode and I will tell you honestly whether your graphics are helping. Sometimes the answer is that they are fine and the problem is somewhere else entirely.