Offline voice (Whisper)

Offline Whisper is a second dictation engine that runs entirely on your phone. Instead of streaming audio to a speech-recognition service, it records what you say and transcribes it locally with an OpenAI Whisper model, so nothing ever leaves the device.

Recognition engine choice on the Voice typing screen.

Why offline dictation

The system recognizer (covered in Voice typing) is fast and shows words as you speak, but it depends on your OS, its installed languages, and (unless the on-device language pack is installed) sending audio to a recognizer service. Offline Whisper trades that for a hard privacy guarantee: it works across many languages and never sends audio anywhere.

The trade-off shows up as a real UX difference, not just a privacy footnote: Whisper isn’t streaming. It records into a buffer and transcribes the whole clip only once you stop talking (or hit the 30-second cap; see Details & edge cases), so the panel shows “Listening. Press when you are done.” instead of live partial words.

Bigger models trade download size and speed for accuracy:

• Tiny: fastest multilingual model. Fine for clear speech, struggles with noise.
• Base: balanced accuracy and speed.
• Small (recommended): the best accuracy that still runs comfortably on a phone, noticeably better on accents and background noise than Base.
• Medium: near-desktop accuracy. Expect several seconds per phrase and close to a gigabyte of RAM while it runs.
• Turbo: large-v3-turbo, so it’s about as accurate as Large at half the download, and much better than the small models at working out which language you’re speaking. An 820 MB download. It was trained for transcription only, so translation isn’t offered for it.
• Large: the most accurate graphs there are, and far too big for most phones. A 1.5 GB download that only high-memory devices can load. Two are listed, and if you’re going to spend the 1.5 GB, Large v3 is the one to take: it’s the newer model and it recognises languages much more reliably than the original Large.

The model catalog

The catalog has 29 downloadable models: 10 multilingual graphs (7 that auto-detect the spoken language, and 3 grouped models you can pin to one language) plus 19 single-language models covering 13 languages (English, German, Spanish, French, Hindi, Italian, Japanese, Portuguese, Russian, Tamil, Telugu, Urdu, and Chinese, several with more than one size). A single-language model only appears in the catalog once you’ve enabled that language on the Languages screen.

Model catalog with size and language filters, grouped by tier.

Turbo and Large v3 need a 128-band mel spectrogram where every older Whisper model needs 80, so each one downloads a different filterbank file alongside the graph and the audio pipeline reads the band count from that file. If the two files ever disagree, dictation fails with a message instead of decoding noise into confident-looking words.

When a language can only be guessed at

Every model has to decide which language you’re speaking, unless it was built for one language or belongs to a grouped model that takes the language as an input. The grouped models list 40 and 26 languages, and those lists are compiled into the graph itself, so a language outside them cannot be forced no matter what the app asks for.

For those languages, and Bangla is one of them, a model can decide you’re speaking a bigger neighbouring language and hand back fluent sentences in the wrong one. Nothing errors, because as far as the model is concerned it did its job. Two things help:

• The model-per-language list says so on any affected language, and points at Turbo and Large v3, which get the language right far more often than the small models do.
• When the wrong language happens to be written in a different alphabet, WM Keyboard converts it back. Bangla read as Hindi is the common case: the model still transcribed the Bangla words it heard, it just spelled them in Devanagari, and the two alphabets line up closely enough to map across. You get readable Bangla instead of a line you can’t read at all. It can’t rescue a genuinely wrong language, only a wrong alphabet.

Whisper itself has an “initial prompt” that biases spelling and vocabulary, and its own way of forcing a language. Neither is available here. The community .tflite conversions this app uses were exported with the entire generation loop traced inside the graph, so the only inputs left are the spectrogram and, on the grouped models, a number selecting one of the languages it was built with.

Picking a model per language

Rather than one global model for every language, WM Keyboard resolves a model per enabled language and falls back in this order:

1. A model you’ve explicitly pinned to that language, if it’s downloaded and covers it.
2. The largest downloaded single-language model for that language.
3. The app’s own fallback model, if it’s downloaded and covers the language. You don’t pick this one; see Options.
4. The best-ranked downloaded model that covers the language at all.
5. As a last resort, that same fallback or the first downloaded model, even if it can’t actually transcribe that language.

That last step is deliberate: it produces wrong words rather than a dead microphone. When a language’s assigned model can’t cover it, the row is flagged in the error color with a note to download one that does.

Model-per-language routing list, with a coverage warning on one row.

Downloading and managing models

Each catalog row carries its own Download button on the right. Tapping the row itself does something else: it expands the row to show what the model is and which languages it covers, so you can read up before spending the bandwidth. Once a model is on disk that button becomes a delete icon; a failed attempt offers Retry and an interrupted one offers Resume.

Each model downloads as two files (the .tflite graph and a companion vocabulary/filterbank file), with the smaller vocabulary file fetched first so a broken connection fails fast. Downloads are resumable: if you lose connection or leave the app, the next attempt picks up where it left off instead of restarting.

Only one model downloads at a time. Before it starts, WM Keyboard checks that you have enough free space (the remaining bytes plus a safety margin); if you don’t, the download stops with a message telling you roughly how much space to free up. On a metered connection, any model at or above roughly 150 MB asks you to confirm before using mobile data.

A model mid-download, with progress and the storage summary below the list.

Downloaded models live in the app’s private storage, not shared storage: the screen shows total space used and lets you delete a model to reclaim it. Deleting a model also clears any per-language pin that pointed at it, and clears the app’s own fallback if that was the deleted model. If an app update ever drops a model from the catalog, leftover files show up as reclaimable space with a one-tap “Free up” button.

When exactly one model is on disk, the app quietly adopts it as that fallback, so dictation works right after your first download with no setup step in between. It does the same again if you delete whichever model was holding the job.

Options

WM Keyboard›Tools›Voice typing

The Whisper-specific controls live inside the same Voice typing screen as the system-recognizer settings: there’s no separate settings destination. The Recognition engine picker below appears Full edition only (it’s what lets you switch to Whisper in the first place); the rest of the table (translation and model routing) only appears once you’ve already selected the offline engine there.

Setting	Default	What it does
Recognition engine	System recognizer	Switches dictation between the system speech recognizer and offline Whisper. Hidden entirely on the lite edition: there’s no disabled option, dictation is always the system recognizer.
Translate to English	Off	Speaks any language and types its English translation instead of a transcription, when the routed model supports it.
Model per language	Automatic	Pins a specific downloaded model to a language, overriding automatic routing. One row per enabled language, each with a Change button.

There’s deliberately no “use this model for everything” picker. Dictation resolves a model from the language of the active layout, the same way the system recognizer follows the layout’s locale, so the only choice worth offering is the per-language one. The fallback that step 3 of the routing order refers to is something the app keeps for itself: it adopts a model when that model is the only one on disk, and clears it if you delete that model. You never set it by hand.

Nine of the 10 multilingual models support translation; Turbo doesn’t, because large-v3-turbo was trained for transcription only, and none of the 19 single-language models do. If a language routes to one of those, the translate toggle has no visible effect for that language.

Details & edge cases

Every clip is a fixed 30-second window. Whisper always processes a full 30-second spectrogram: a short utterance is padded, and hitting the 30-second mark auto-triggers transcription of whatever was captured. This makes transcription latency roughly constant per phrase rather than scaling with how long you spoke.

Only two rows here can be reset. Recognition engine and Translate to English are ordinary settings, so each grows the usual restore button once you move it off the default; see Putting one setting back. The per-language routing rows and the catalog aren’t settings rows at all, so they have no button. You undo a pin by opening that language’s dialog and picking Automatic again.

Speed has no benchmarked numbers. There’s no measured per-device timing table in the app; the only official speed guidance is the relative sizing above (Tiny fastest, the two Large graphs slowest and heaviest on RAM) plus a general note that bigger models are more accurate but slower to both download and transcribe.

Full edition only. Offline Whisper is unavailable on the lite edition: the “Recognition engine” section doesn’t render at all, so dictation always uses the system recognizer there. See Full vs Lite if you’re not sure which one you have.

Power saving can silently override your choice. When power saving is active and its “On-device models” option is on (which is the default, and which arms automatically whenever Android’s own battery saver turns on), dictation falls back to the system recognizer even if Whisper is selected and a model is downloaded. If dictation seems to have quietly switched engines, check your power-saving settings first.

No mic opens without a model. If Whisper is selected but nothing is downloaded yet, the panel says so and offers both ways forward instead of opening the microphone: one chip opens this screen to download a model, the other switches back to the system recognizer and starts listening right away. That switch is saved, the same as changing the engine in settings. The compact bar has room for one line, so it points you at settings, where both choices live.

Panel keys don’t end a recording. Space, backspace and enter on the voice panel’s right-hand rail edit the text while the microphone stays open, so you can add a space or fix a character without losing what you’ve said. The same goes for a keystroke that lands while a clip is being transcribed: the audio is already captured, so the words still arrive, they just land after whatever you typed. Typing on the keyboard itself during a system-recognizer dictation still ends that utterance, because its live partial text sits in the editor and a keystroke inside it would corrupt it.

Privacy and downloads are separate concerns. Transcription itself makes no network calls. The only network activity Whisper causes is fetching a model file over HTTPS from its public repository when you tap download (see Network policy for the full picture of what WM Keyboard can send and when).

Attribution. The engine is TensorFlow Lite, branded LiteRT (Apache-2.0, Google); the model architecture and weights are OpenAI’s Whisper (MIT); the .tflite graphs themselves are third-party conversions hosted on Hugging Face, not built by this project. Full credits are in the app’s About screen.

Related: Voice typing covers starting dictation, strip vs. panel mode, continuous mode, and spoken punctuation, all shared behavior between the system recognizer and Whisper.