Caption Flow Get the app
An illustrated economics lecture with a chart on the board

Offline or online captions: which should you use?

Caption Flow gives you three ways to turn speech into captions: offline models that run on your phone, Google captioning, and AI captioning. They all show up in the same caption box, but they make different trade-offs between privacy, accuracy, speed and cost. Here’s how to choose.

The short answer

  • No connection, or you want nothing to leave your phone: offline.
  • A specific language, especially one offline models don’t cover: Google captioning.
  • Several speakers, mixed languages, or you need the best accuracy: AI captioning.

Offline captioning

Offline captioning uses a speech model you download once. After that, everything happens on your phone: the audio is never sent anywhere, and captions work on a plane, in the underground, or with mobile data off.

You can pick the model that suits your phone and patience:

  • Fastest shows words with the least delay, at lower accuracy.
  • Balanced is the middle ground for most phones.
  • Higher accuracy makes fewer mistakes, but takes a little longer and works the processor harder.

Offline models cover a set of major languages, including English, Spanish, French, German, Russian and Chinese. They’re free to use.

Good for: privacy-sensitive conversations, travel, saving data, and long sessions where you don’t want to use up online time.

Google captioning

Google captioning sends audio securely to Google’s speech service and returns captions in real time. It’s noticeably faster and more accurate than offline models, and it covers more than 140 languages. You choose the language being spoken from a list.

When you’re signed in, the free plan includes a monthly allowance of Google captioning. Premium plans add more: 10 hours a month on Go, 20 on Plus, and unlimited on Pro.

Good for: one known language, clear audio, and languages the offline models don’t cover.

AI captioning

AI captioning is Caption Flow’s most accurate option. Beyond writing down words, it:

  • detects the spoken language automatically, so you don’t have to pick it, even when people switch languages;
  • labels different speakers, so a transcript of a meeting shows who said what;
  • copes with people talking over each other;
  • translates as it captions, if you turn translation on.

AI captioning is part of Premium and runs on monthly credits: roughly 5 hours on Go, 10 on Plus and 30 on Pro. Those hours are an estimate, since credits depend on how much is said, so very fast talking uses them faster. AI translation uses some extra credits.

Good for: meetings, interviews, lectures with questions from the room, multilingual videos, and anyone who relies on captions to follow a conversation.

Side by side

OfflineGoogleAI
Needs a connectionNoYesYes
Audio leaves the phoneNeverSent to Google, not stored by usSent securely, not stored by us
LanguagesMajor languages140+, you chooseDetected automatically
Speaker labelsNoNoYes
PlanFreeFree allowance, more with PremiumPremium

Switching is easy

You don’t have to commit to one. In Live Caption’s audio language screen, the Offline and Online tabs sit side by side; change the engine before you start a session. Many people use offline captioning day to day and switch to AI captioning for the meetings and lectures that matter.

Whichever you choose, Caption Flow doesn’t store your audio, and transcripts are kept as text on your phone only. See the privacy policy for details, or get Caption Flow on Google Play to try all three.