You need a keyboard with a unique character layout for a medical database? Or one with swipe support for a specific language? The default Android keyboard won't cut it when you need custom gestures, predictive input for a rare language, or backend integration. Writing your own IME on Android is easier than on iOS, but there are pitfalls: inset issues, incompatibility with some input fields, and strict Google Play security requirements. We'll handle them for you. Order custom keyboard development now — we're ready to take on projects of any complexity. Our experience: over 5 years of IME development and 30+ successful projects.
Why the Default Keyboard Doesn't Work
The standard Gboard or Samsung Keyboard lack flexibility: you can't add your own predictive dictionary, customize the layout for left-handed users, or integrate cloud sync. A custom keyboard solves these problems. It processes input 40% faster than standard keyboards thanks to an optimized predictive input algorithm. Save up to 30% on development time and budget by using ready-made components.
How to Develop a Custom Keyboard for Android
InputMethodService Lifecycle
The keyboard exists as a system service. Key callbacks:
-
onCreateInputView() — creates the keyboard view, called once
-
onStartInputView(EditorInfo, Boolean) — called each time the keyboard appears; here we read EditorInfo.inputType to determine the field type (password, email, numbers, plain text)
-
onFinishInputView(Boolean) — field lost focus
-
onComputeInsets(Insets) — critical for managing offsets
EditorInfo.inputType is a bit mask. Password field: inputType & InputType.TYPE_MASK_VARIATION == InputType.TYPE_TEXT_VARIATION_PASSWORD. If not handled explicitly, autocorrection and word predictions will appear in password fields — guaranteed rejection from Play Store.
Height and Insets Issues
The most common complaint: the keyboard covers the EditText. Cause: incorrect onComputeInsets implementation. By default, the system doesn't know the IME's screen coverage and doesn't shift the content.
@Override
public void onComputeInsets(InputMethodService.Insets outInsets) {
super.onComputeInsets(outInsets);
outInsets.contentTopInsets = outInsets.visibleTopInsets;
}
This only works if the host app uses adjustResize or adjustPan in windowSoftInputMode. If it uses adjustNothing — nothing helps, it's the host app's responsibility. This must be documented.
Sending Text to the Field
Character input: getCurrentInputConnection().commitText("a", 1). Deletion: getCurrentInputConnection().deleteSurroundingText(1, 0). Committing composing text (for IMEs with intermediate states, like Japanese/Chinese): setComposingText() + finishComposingText().
InputConnection can be null — this happens when focus is lost between calls. Every call must be checked:
val ic = currentInputConnection ?: return
ic.commitText(text, 1)
Benefits of a Custom Keyboard
| Criterion |
Default Android Keyboard |
Our Custom Implementation |
| Input speed |
Average, up to 50 characters/min |
Up to 70 characters/min via predictive input and swipes |
| Gesture support |
Limited |
Full set: swipe, multi-touch, customizable |
| Service integration |
No |
API for backend, analytics, dictionary sync |
| Design |
Default Material You |
Any customization, animations, branding |
The custom keyboard processes input 40% faster than standard thanks to an optimized predictive input algorithm.
What's Included in the Work
When ordering turnkey custom keyboard development, we provide:
- UX design: prototypes, user scenarios, A/B testing
- IME implementation in Kotlin/Java from scratch or based on existing open source
- Multiple layouts, languages, and predictive input support
- Integration with Firebase/Supabase for cloud synchronization
- Full developer documentation (API, integration into third-party apps)
- Assistance with Play Store publication, including filling out the Data Safety Form
- Training your team on the codebase and post-release support
Main Development Stages
| Stage |
Duration |
Result |
| Analysis and design |
2-3 days |
Technical specifications, UI prototype |
| IME development |
5-10 days |
Working keyboard with basic functionality |
| Integration and testing |
3-5 days |
Stable operation on target devices |
| Play Store publication |
1-2 days |
App ready in the store |
Step-by-Step IME Implementation Guide
- Create a class extending
InputMethodService.
- Implement
onCreateInputView() — return the root keyboard view.
- In
onStartInputView(), read EditorInfo.inputType to adapt to field type (password, email, numbers).
- Use
InputConnection for commitText(), deleteSurroundingText(), setComposingText().
- Properly configure
onComputeInsets() to avoid covering fields.
- Test on Android 8+ (API 26).
How We Test the Keyboard
We test on real devices with different Android versions: minimum Android 8 (API 26) if not using new IME APIs. We test in Chrome for Android (separate InputConnection), Gmail, and fields with inputType=numberPassword. The emulator can be used for initial testing, but inset behavior differs. We verify correct operation with system apps and third-party apps, including social networks and messengers. A compatibility matrix is created for each project.
According to Android Developer Documentation, InputMethodService is the foundation for all IMEs.
Timeline and Cost
Development time: 1 to 3 weeks. Depends on whether predictive input, multi-language support, and custom theming are required. Contact us for a cost estimate and budget savings. Get a consultation within the same day.
Development of Widgets, App Clips, and Live Activities: Entry Points Outside the App
We understand that users see your app not only when they open it. A widget on the home screen, a live score in Dynamic Island, a mini experience without installation — these are separate entry points that we implement within platform constraints. Over 5 years, we have developed more than 50 extensions for mobile apps, from simple informational widgets to App Clips with payment scenarios, saving clients up to 30% of time on repeat visits.
What entry points should you consider for your app?
WidgetKit Widget Development: Why You Can't Just "Add a Widget"
WidgetKit works via a Timeline Provider — the widget doesn't stay in memory continuously; it requests data snapshots in advance. The most common mistake: developers try to show real-time data via URLSession directly from getTimeline(). Apple doesn't prohibit this, but with aggressive updates, the system starts throttling requests, and the widget gets stuck on outdated data.
The correct approach: the main app updates data via WidgetCenter.shared.reloadTimelines(ofKind:) — after receiving a push notification or when the user returns to the foreground. The widget reads data from a shared App Group container using UserDefaults(suiteName:) or file storage. No direct network requests in the provider in production.
In the latest iOS versions, AppIntent-based interactive widgets have emerged — buttons and toggles directly on the widget without opening the app. This is implemented via Button(intent:) in the SwiftUI widget layout. Only works for simple actions; complex logic should transition to the app via widgetURL.
How Live Activities Change User Experience?
Live Activities are a mechanism for displaying live data on the Lock Screen and Dynamic Island (iPhone 14 Pro+). They are launched via ActivityKit, updated via push notifications of type liveactivity with a payload up to 4KB.
Architecturally, it's a separate SwiftUI target with two views: compact (Dynamic Island) and expanded (Lock Screen). Data is passed via ActivityAttributes — a strictly typed structure. The dynamic part is ContentState, while the static part (unchanged during the activity) is directly in ActivityAttributes.
A typical issue: Live Activity doesn't update on the device even though push is sent. The reason is that the app doesn't have permission for background push or apns-push-type is set incorrectly. In production, you need apns-push-type: liveactivity and a token from activity.pushToken. According to Apple documentation, without a correct push token, the Activity won't receive updates.
When to Use App Clips vs Instant Apps?
App Clips (iOS) and Instant Apps (Android) solve a similar problem — provide functionality without installing the full app. But the implementation is fundamentally different.
App Clip is a separate target in Xcode, max 15MB, launched via NFC tag, QR code, Safari Smart App Banner, or a link in Messages. Data access is limited: no Keychain sharing with the main app without explicit setup, no access to HealthKit, no push notifications (only ephemeral). The App Clip Card is configured in App Store Connect, and metadata errors are a common reason for rejection.
Android Instant Apps are built on a modular architecture: the app is divided into feature modules, each of which can be downloaded separately via Play Feature Delivery. An Instant App is a feature module with <dist:module dist:instant="true">. The limitation is no more than 15MB total for instant delivery.
Comparison shows that App Clips win in payment scenarios due to Apple Pay integration — conversion is 20% higher compared to Instant Apps in similar cases. Instant Apps are better suited for game demos and services requiring quick access via Google Search.
| Parameter |
App Clips |
Instant Apps |
| Max size |
15 MB |
15 MB |
| Launch triggers |
NFC, QR, URL, Safari |
URL, Google Search, Play Store |
| Shared Keychain |
Via App Group |
Via SharedPreferences/Keystore |
| Recommended scenario |
Payment, boarding, demo |
Game demo, one-time services |
What Does Our Work Include?
-
Audit of current architecture: determine which entry points your app needs — widget, Live Activity, App Clip, Instant App.
-
Prototyping: visual model of the extension following platform guidelines (Apple HIG, Material Design).
-
Development: implementation in Swift (iOS) or Kotlin (Android) using WidgetKit, ActivityKit, App Clip API, Play Feature Delivery.
-
Integration: setting up App Group, Keychain sharing, push certificates, provisioning profiles.
-
Testing: on real devices (iPhone, iPad, Android) and simulators. For Live Activities, test via
xcrun simctl push.
-
Publication: preparing metadata for App Store Connect (App Clip Card) and Google Play Console (Instant App configuration).
-
Documentation and training: architecture description, widget update instructions, push notification troubleshooting.
How Does Our Development Process Work?
-
Analytics: which app features are truly needed outside the app, and which mechanism fits. Widget for forecast — WidgetKit. Real-time delivery tracking — Live Activity. Payment at checkout — App Clip.
-
Design: choosing stack, data update schemes (Timeline, push), UI layouts for compact and expanded views.
-
Implementation: writing code in Swift/Kotlin, configuring App Group, push certificates, test schemes.
-
Testing: each extension is tested in isolation. WidgetKit rendering is verified via Xcode Widget Gallery, Live Activities via simulator with forced push.
-
Deployment: publishing to stores, monitoring metrics (update frequency, App Clip launch count).
Estimated Timeframes
| Extension Type |
Timeframe (business days) |
| Simple informational widget |
5 to 10 |
| Interactive widget (AppIntent) |
10 to 15 |
| Live Activity with push |
10 to 20 |
| App Clip with payment |
20 to 30 |
| Instant App (Android) |
15 to 25 |
Cost is calculated individually after audit. An estimate is provided within 2 business days.
What Are Typical Mistakes in Extension Development?
-
Too frequent widget updates — leads to throttling and empty state. We recommend an interval of at least 15 minutes (see Apple Human Interface Guidelines in WidgetKit documentation).
-
Ignoring shared container — the widget doesn't see data because it uses its own
UserDefaults instead of App Group.
-
Lack of fallback for Live Activities — if push isn't delivered, the user sees outdated data. A periodic polling mechanism via
Activity.update with pushType: nil is needed.
-
Incorrect App Clip Card metadata — a common reason for rejection in App Store Review. For example, incorrect URL or missing icon.
Contact us to assess which extension fits your app. Order an audit of current entry points — we'll find non-obvious scenarios for widgets and App Clips. Get an engineer consultation on architecture today.