We got caption fonts wrong. Here's what we learned.
Three weeks after we shipped build 3 of Clipr, a church social media manager from Sheffield sent us a screenshot. The captions on her sermon clips were too small. Not broken, not invisible, but small enough that she had to choose between readability on a phone and looking professional on a screen. She wasn't angry. She was just matter-of-fact: 'Can we make them bigger?' That question sent us back to the drawing board.
The problem we thought we were solving
When we built the caption system for build 3, we had one goal: get auto-captions baked into the output video so creators wouldn't have to fiddle with subtitle overlays in another tool. That was the hard part. We spent weeks getting Apple Speech to transcribe accurately on-device, then stitching those captions directly into the final clip.
We chose the font size based on what we thought looked balanced. A 9:16 vertical clip is small. We reasoned that larger captions would eat too much of the frame. A pastor's face, we thought, had to stay the hero. So we picked a size that felt proportional to us, sitting at our desks on 27-inch monitors, reviewing test clips of sermons and podcasts in isolation.
The Sheffield manager was watching those same clips on a 6-inch screen, on the London Underground, at half brightness. The math was different.
Why it mattered more than we expected
Here's what we missed: the people using Clipr are not uploading to YouTube. They're uploading to TikTok, Reels, Instagram Stories. They're reaching people who scroll for three seconds and move on. If someone can't read your caption in that window, the clip fails silently. There's no 'turn on subtitles' button. There's no second chance.
We also underestimated how many clips get watched without sound. Reels autoplay muted by default. TikTok's algorithm doesn't care whether captions are there, but users do. A pastor's teaching is built on words, not just tone and presence. If those words disappear into the frame because the font is too small, the entire clip's purpose collapses.
The Sheffield manager wasn't being precious about design. She was telling us that our output wasn't fit for the platform it was meant for.
What we actually changed
We increased the default caption size by 40 per cent and tested it again. On phones. On TikTok. On Reels. Not on our desks.
But we didn't just make them bigger. We also changed the font weight. The original was too thin against the video background. We moved to something bolder, with better contrast. We also shifted the placement slightly lower to avoid sitting on top of faces in shots where someone was standing close to the camera.
None of this was sophisticated. We were fixing an obvious mistake, dressed up as a design choice. The real lesson wasn't about captions, though. It was about testing where the product lives, not where we work.
The harder conversation
What bothered us more was that we'd shipped this without asking. Clipr has a Creator tier and a Pro tier. Captions come baked in on Creator, which means a lot of our users never see them until they're already on TikTok. By then, it's too late to change.
We could have sent a test clip to five pastors we know before build 3 went live. We didn't. We were moving fast and we trusted our instinct. The instinct was wrong.
This is the tension in building tools for creators. You want the output to be beautiful and ready to ship. You also want to stay out of the way. Big captions feel intrusive if you're a minimalist. Small captions feel useless if you're trying to reach people scrolling at speed. There's no universal answer.
What we should have done was ship with options. Let the creator choose. We're still thinking about how to do that without adding noise to the interface, but we're committed to it. Build 4 will have caption size choices for Creator and Pro.
Why this matters for what's next
Clipr is built for a specific job: turning long sermons and podcasts into short clips. The people using it are not video professionals. They're busy. They're using their phones. They want something ready to post, not another tool to learn.
That constraint is good. It keeps us honest. Every choice we make has to survive the question: 'Does this help a church social media manager post a clip on Tuesday morning before the service starts?'
The caption font doesn't seem like it matters until you're the person in Sheffield waiting for the video to load, squinting at small white letters, wondering whether your congregation will actually read them. Then it matters a lot.
We're releasing build 3.1 next week with the larger, bolder captions as standard, and build 4 will add caption size choices for anyone who prefers something different. But the real question isn't about captions. It's this: what else are we testing at our desks that only breaks when it meets reality?
Ready to try Clipr by MRVL?
One tap to download. No sign-up wall.