Loquela
Research

Android voice keyboards take different approaches

Gboard, FUTO, Wispr Flow, and Loquela make different choices about integration, offline processing, model control, and rewriting.

4 min readResearch checked 2026-11-13

Voice typing on Android is not one product category with one implementation. The microphone may belong to the keyboard, a system speech service, a separate voice-input app, or a cross-platform AI dictation service.

The public documentation for Gboard, FUTO, Wispr Flow, and Loquela shows four approaches.

ProductPublicly documented approachMain user-facing control
GboardGoogle's advanced voice typing runs on Pixel 6 and newer, with voice commands and automatic punctuation. Google says the text you speak stays on your device.Tight integration with Google's keyboard and supported Pixel features.
FUTOFUTO Keyboard has offline voice input built in, and FUTO Voice Input remains a separate app for other keyboards. FUTO says its voice input uses no internet connection.Offline voice input in FUTO's own keyboard, or as a service for compatible keyboards.
Wispr FlowOn Android, a floating button that works alongside your keyboard. Its privacy page says transcription always happens in the cloud, and Flow cleans up filler words and spoken corrections.Automatic cleanup and writing-oriented interpretation.
LoquelaA complete Android keyboard with selectable local speech models, optional transcription through your own third-party API key, and no AI rewriting pass.Model choice, explicit processing location, and the ability to turn voice off entirely.

Sources: Gboard advanced voice typing, FUTO Keyboard, FUTO Voice Input, Wispr Flow privacy, and Wispr Flow on Android. Product documentation was checked on November 13, 2026. The first version of this post was checked on July 31, 2026.

Integration changes the experience

Gboard can combine keyboard behavior with device-specific Google features. FUTO builds voice into its own keyboard and also offers it as a separate service that other keyboards can call. Wispr Flow adds a dictation button on top of whatever keyboard you already use and carries the same service across several platforms. Loquela owns both the keyboard and the local model-management path.

None of those structures is automatically correct for every person. A separate speech service can work with multiple keyboards. A deeply integrated keyboard can coordinate more state. A cross-platform service can offer a consistent account experience. A local model catalog can expose control that a managed service keeps internal.

Offline is not one switch

FUTO and Loquela both make offline recognition a central part of their public design. Their packaging differs. FUTO builds offline voice into FUTO Keyboard and keeps FUTO Voice Input for other keyboards. Loquela includes voice inside its keyboard and lets you choose among 13 verified voice models.

Gboard's advanced voice typing on Pixel keeps the text you speak on the device, though Google notes that some editing features send text to its servers. Its public documentation does not expose a general speech-model chooser comparable to Loquela's catalog.

Wispr Flow says transcription always happens in the cloud, and it emphasizes cleanup: removing filler words and following spoken corrections. That is a different promise from direct local transcription. It may be attractive when polished text is the goal and less attractive when you want the exact spoken wording or an offline path. Loquela and Wispr Flow on Android compares the two in more detail.

Model choice has a cost

Exposing models creates setup decisions. Users have to understand storage, memory, language coverage, speed, and context. Loquela addresses that with device filtering, recommendations, practical labels, and a model finder.

A managed service can hide those decisions, which is simpler. It also means the service chooses the model and update path.

The useful comparison question

The useful question is not which keyboard wins every row. It is which tradeoff the user wants:

  • Deep platform integration
  • A separate offline speech service
  • AI-assisted rewriting across devices
  • A full keyboard with local model and processing control

Loquela is built around the fourth choice while retaining normal tap typing, autocorrect, editing, themes, clipboard tools, and Android inline autofill. Voice is central to the product, but it can be skipped or disabled without removing the rest of the keyboard.