The hidden cost of caption churn in multilingual meetings

Voice translation tools are transforming how global businesses collaborate and share knowledge across borders. When deployed effectively, real-time on-screen captions enable more inclusive, productive meetings where alignment is faster and conversations are more meaningful.

Why? Because everyone can contribute in the language they’re most comfortable in and be instantly understood. Knowledge flows more naturally, Q&As happen spontaneously, and experts can join conversations regardless of the language they speak.

All of this is powerful. But the real question is, how do these tools hold up in real-world business scenarios? When the stakes are high, accuracy and reliability are non-negotiable.

Translation quality is often the first thing that comes to mind. But what truly affects usability in live meetings are unstable meeting captions.

Because what’s the point of an accurate translation if you can’t read it in real time? Constant flickering and rewrites make captions hard to follow and disrupt the flow of conversation. That’s why eliminating caption churn is so key to unlocking productive multilingual meetings.

The impact on your business

Imagine investing in something as transformative as voice translation, only to spend more time correcting mistakes and rebuilding trust. That’s the risk of high caption churn. 

The more captions flicker and change, the more distracting they become. Reduced readability increases cognitive load, making it harder to absorb information and stay present in the conversation. The result is confusion, slower decision-making, and reduced efficiency.

In high-stakes meetings, this can lead to costly misunderstandings, misalignment, and errors.

Despite its clear impact on the efficacy of voice translation tools, caption stability has rarely been measured in independent, comparative studies. Until now.

The Slator study: bridging the gap

Slator conducted an independent evaluation of AI-generated captions across five leading platforms: Google Meet, Microsoft Teams, Zoom, DeepL Voice for Microsoft Teams, and DeepL Voice for Zoom.

A total of 28 professional linguists carried out a blind evaluation of translated on-screen captions across 14 language combinations. The study assessed translation quality, caption stability, and error rates using real-world, conversational audio samples.

Slator measured the frequency of “change events” across languages and platforms, including how captions evolve as a sentence builds, how often previously displayed text is revised, and instances of flickering before stabilizing.

These findings were consolidated into a single 0–100 score reflecting both translation quality and caption stability.

And the results revealed a clear winner. 

DeepL Voice leads in both stability and accuracy

DeepL got the highest caption stability score in Slator’s independent evaluation, achieving 88.6/100. Across all language pairs, it reduced churn by 37.6% compared to Microsoft Teams and 54.7% compared to Zoom. 

Most importantly, this stability does not come at the expense of accuracy. 

In a combined score measuring both caption stability and translation accuracy, DeepL Voice achieved the highest result of 96.4/100, compared to an average of 87-89 for competing platforms. 

DeepL Voice also reduced the rate of critical errors by 76% on average—making it particularly well suited to high-stakes settings where accuracy and nuance are non-negotiable. And 96% of professional linguists ranked it their preferred platform for translated captions overall.

Let your ideas do the talking

Caption stability isn’t just a nice-to-have. When it comes to high-stakes meetings, it can spell the difference between a conversation that drives results and one that leads to costly mistakes and reputational damage.

The last thing you need in the boardroom are flickering, shaky, unstable meeting captions that aren’t just difficult to read, but set the wrong tone and impression from the start. Because stable, high-quality captions reflect a stable, high-quality business. They should fade into the background and let your ideas, and expertise do the talking, not become a distraction your team has to focus on.

DeepL Voice enables exactly that for Microsoft Teams and Zoom Meetings, in 100+ languages. Powered by specialized AI trained on proprietary data by thousands of language experts, it delivers unparalleled accuracy and nuance in every business conversation. 

Read the full Slator report and explore the complete findings here.