Sources checked October 21, 2026 · Not hands-on tested
Dictate Keyboard: Android Setup, API Costs & Offline Limits
Dictate Keyboard: Voice Typing is DevEmperor’s Android app, package net.devemperor.dictate. Looking for Dictate’s older “Whisper AI keyboard” descriptions? Check this developer and package. Current setup offers your own API key, a local speech model, your own server, or prepaid cloud credit. The app purchase and cloud usage costs are separate.
- Android · DevEmperor
- U.S. Play price: $4.49
- Offline recognition requires a model
Why older setup guides can mislead
The current app is a full keyboard built on FlorisBoard, replacing the Java app used for versions 1–3. We cross-checked the released 6.4.0 release, dated October 7, against the store changelog. A tutorial requiring an API key and internet for every dictation does not describe all current choices.
Choose a setup and understand the bill
These four routes are documented in the released-version README. Choose based on where you want processing to happen and how much configuration you are willing to maintain.
| Route | What you configure | Additional cost | Processing boundary |
|---|---|---|---|
| Your own API key | Choose a provider and enter your credential in the app. | Provider usage charges or account-specific allowance. | Audio goes to that provider; its terms apply. |
| On-device recognition | Download a compatible speech model before going offline. | No speech-provider usage bill for local recognition. | Speech recognition stays on the phone; check rewriting separately. |
| Your own server | Configure a compatible endpoint and its model. | Your hardware or hosting costs. | Audio reaches that server; self-hosted does not mean on-phone. |
| Dictate Cloud credit | Buy optional prepaid credit through Google Play. | Credit purchases in addition to the app download. | Requests pass through the developer’s cloud service. |
Google Play showed $4.49 with in-app purchases when checked; your regional checkout may differ. Open source does not make that store download free. Release 6.4.0 has no APK attached: building from source is a separate technical route. For cloud use, check your provider’s billing or the in-app credit pack before recording a long session.
Android setup: start with one short draft
Follow the installed app’s onboarding if labels differ. This sequence combines its documented choices with a suggested check in your own apps.
- Install the DevEmperor app linked above. Open onboarding, enable Dictate in Android’s keyboard settings, then select it as your input method.
- Choose a processing route. For local recognition, open Settings → AI providers → On-device (offline) and finish the model download while connected.
- For bring-your-own-key use, select a provider and enter your own credential only in the installed app. Check billing and choose the transcription model. Prepaid credit is an alternative, not a required purchase.
- Allow microphone access for dictation. Leave the optional floating button off unless you need it; its Accessibility permission is separate from using the keyboard.
- Open a normal text field in a notes app, dictate a harmless sentence, and check names and numbers. Test rewriting separately before using the setup for important messages.
Offline speech is only one part of privacy
The developer’s versioned privacy policy distinguishes local recognition, direct provider requests and optional Dictate Cloud. With your own key, audio goes directly to your selected provider, whose retention rules apply. Dictate Cloud adds the developer’s server; the policy says audio and text are not retained there, while usage and purchase records are kept separately.
Dictation history is on by default and stored locally. Audio history is a separate option, with failed-request exceptions. Review history, backup and provider settings rather than reading “offline” as “nothing is ever saved.” A later cloud rewrite can send text off-device even after local transcription. We have not audited network traffic or the installed app.
Our suggested offline check: download a speech model, select it, disconnect, and repeat a short dictation. Then test your desired rewrite separately. A successful transcript does not verify the second stage, and your phone’s memory and chosen language still matter.
Who should consider it?
Consider Dictate when choosing your speech engine or maintaining a local/server setup matters to you. Start with a simpler built-in option if you mainly want a microphone that works immediately. Compare correction effort on your own messages; we have not measured accuracy, latency or battery use, and do not name a performance winner.