Sound-On vs Sound-Off Ads: How to Design for Both
How to design video ads that work whether the viewer has sound on or off, and why that matters.
Design your ad to work with the sound off first, then treat sound as an enhancement rather than a requirement. Most feed and Reels placements autoplay muted, so if your message only lands through audio, a large share of viewers never get it. Burned-in captions, clear on-screen text, and a visually obvious hook let the ad work either way, and that combination consistently outperforms an ad that depends on the viewer tapping to unmute.
Why does sound-off matter so much?
Autoplay in most feed environments starts muted by default, and a large share of viewers scroll past without ever tapping to unmute. If your hook or offer only makes sense with audio — a voiceover explaining the product, a sound effect timed to a reveal — you lose that audience entirely in the first second. Designing for sound-off isn't pessimistic, it's just matching how the placement actually behaves for most viewers.
Does that mean sound doesn't matter at all?
No — sound still adds real value for the viewers who do have it on or who unmute after the hook catches them visually. Voiceover, music, and sound effects can deepen the pitch, add urgency, or make a UGC-style ad feel more authentic and less like a polished commercial. The goal is layering: the visual and captions carry the core message alone, and sound adds texture and persuasion on top for anyone listening.
What does a sound-off-safe ad actually look like?
Burned-in captions that track the spoken line, on-screen text for the offer or key benefit, and a visual hook in the opening frame that doesn't depend on narration to land. Keep captions and text inside the safe zones so platform UI doesn't cover them, and make sure any critical demonstration is visually clear on its own, not just described in voiceover.
How do you produce both a silent-safe cut and a full-audio version without doubling the work?
This is usually a production bottleneck more than a creative one — captioning and voiceover normally mean extra tools and extra time. Polaris folds both into the same workflow: you ask for the ad in a Claude chat, and one-click burned-in captions and AI voiceover (via ElevenLabs) are part of the same generation, so the sound-off and sound-on versions of the ad are the same asset rather than two separate builds. A live panel renders the finished ad in the conversation, and failed renders auto-refund, so testing extra variations doesn't carry the same cost risk as reshoots would.
How do you tell which approach is working?
Track hook rate to see whether the visual-first opening is holding attention on its own, and connect results to ROAS to confirm captions and voiceover are actually improving outcomes, not just watch time. A hook generator can help draft an opening that reads clearly with the sound off, and a ROAS calculator helps you weigh whether the extra audio layer is worth keeping in future batches.