Pro Features

BiBi Keyboard Pro builds on the OSS version with more advanced capabilities for power users.

Overview

Pro includes these exclusive features, grouped in four categories that match the sections below:

Recognition & recording

• Hotwords management: provider-aware hotword adaptation (inject into engine, post-recognition similar-word replacement, polish prompt injection) to improve recognition for proper nouns
• Continuous speaking mode: auto start/stop recording for hands-free multi-segment dictation
• Input field context: AI post-processing can reference nearby text around the cursor for better continuity

Text processing

• Offline Traditional Chinese conversion: convert results to Traditional Chinese automatically
• Regex post-processing: apply regex rules for text post-processing
• App-specific prompts: automatically switch AI prompt preset by foreground app
• AI Assistant: trigger voice commands with wake words, preset keywords, and fuzzy matching

Personalization

• Advanced UI theming: more color customization options
• Omni-direction cursor sliding: enhanced cursor control with 4-direction sliding

Data

• WebDAV auto backup: scheduled cloud backup across devices (see Backup and Sync)

Hotwords Management Pro

Manage custom hotwords to improve recognition of proper nouns, brand names, names, etc.

How it takes effect

Hotwords have three combinable switches:

• Inject into the recognition engine (Inject hotwords into recognition engines): hotwords participate in recognition according to provider support; selected local models support them too
• Post-recognition similar-word replacement (Replace similar words after recognition): after recognition, similar-sounding fragments are replaced with the target hotword via phoneme matching
• Polish prompt injection (Add hotwords to the AI polishing prompt): during AI polishing, enabled hotwords and their aliases are appended to the prompt as a structured word list, so the model can restore misheard proper nouns; independent of post-recognition replacement and may increase token usage. The sub-switch Only include phonetically related hotwords injects only hotwords phonetically close to the current transcript (Chinese and English only) to keep the prompt short

Combinations of the first two switches:

Inject hotwords	Replace after recognition	Result
On	Off	Hotwords participate according to provider support; no pronunciation fallback runs afterward
On	On	Hotwords participate in recognition, then targets and aliases provide a pronunciation-based fallback
Off	On	Recognition engines receive no hotwords, but post-recognition replacement still runs
Off	Off	Hotwords are not applied before or after recognition

Inject hotwords into recognition engines is enabled by default. Actual support and quality still depend on the selected provider.

Other capabilities:

• Target word + aliases: each hotword has up to 3 alias slots. The target word always occupies the first alias slot, and you can add 2 extra aliases
• Aliases participate in phoneme matching: both the target word and extra aliases are matched after recognition; matches are replaced with the target word
• Hotword stats: view trigger frequency and hit stats to iteratively tune your hotword list
• Linked with AI Assistant: AI Assistant keywords can be auto-synced into the hotword list to reduce duplicate maintenance

How to use:

1. Open BiBi Keyboard Pro settings
2. Go to Settings → Smart → ASR Settings → Result Optimization (Pro)
3. Enable Inject hotwords into recognition engines so hotwords participate according to provider support
4. Open Hotword management to add hotwords or batch-import them from the clipboard
5. Long-press a hotword chip to edit aliases (target | alias 2 | alias 3)
6. Enable Replace similar words after recognition and/or Add hotwords to the AI polishing prompt as needed
7. Review trigger frequency/hit stats on the management page and refine your list

Avoid too many hotwords. Per-provider limits apply and excessive lists may hurt performance.

Learn hotwords from corrections

When enabled, Pro compares a committed voice result with your subsequent correction and tries to extract content suitable for the hotword list.

1. Open Settings → Smart → ASR Settings → Result Optimization (Pro)
2. Enable Learn hotwords from corrections
3. Optionally enable Use an LLM to analyze corrections; this requires a working LLM configuration
4. Dictate through modified Fcitx5 / Trime, the latest IME Bridge, or an accessibility-assisted floating ball, then correct a recognition mistake
5. Return to “Learned hotwords awaiting confirmation” to approve or remove candidates

By default, a new candidate waits for manual confirmation; learning the same correction again confirms it automatically. If both Automatically adopt learning results and Use an LLM to analyze corrections are enabled, an LLM-approved candidate can be added after the first correction. Candidates found by local rules still require a second occurrence or manual confirmation.

Companion versions

Modified Fcitx5 / Trime report corrections through optional AIDL function. Other third-party IMEs and the floating ball can use the latest LSPosed/LSPatch IME Bridge module or the floating ball accessibility calling method to trigger hotword learning.

Privacy and false learning

Password, email, URL, phone, and no-personalized-learning fields are excluded. With AI learning enabled, only the changed fragment and a small amount of context are sent to the selected LLM, never the full input field; local rules remain available when the LLM fails; possible false learning, can be combined with manual confirmation to avoid accidentally adding temporary rewrites or common words as hotwords.

Examples and stats

Alias examples:

• Fix similar-sounding words: target word 音素, extra aliases 因素 and 严肃. If the transcript contains 因素, it is replaced with 音素.
• Shortcut phrase input: target word xxxx@qq.com, extra alias primary email. When you say "primary email" and it is recognized, it is replaced with the real email address.

Suggestion

Replace similar words after recognition works best for Chinese and English. For other languages, it may not work properly. It is best for proper nouns and frequent misrecognitions. Adding too many common words may cause unwanted replacements for similar-sounding text.

Convenience:

• Clear all + undo: clear your hotwords in one tap, with an "Undo" action in the snackbar
• Add from selection menu: select any text in any app, then choose "Add hotword to BiBi Keyboard" to add it instantly

Provider support

Provider	Support	Max count
Volcengine	✅	100 tokens (bidirectional streaming) / 5000
DashScope	✅	10000 tokens
Soniox	✅	8000 tokens
SiliconFlow	✅	almost unlimited (SenseVoice/TeleSpeech not supported)
StepAudio	✅	hotword array only; no extra context injection
OpenAI	✅	depends on model
Gemini	✅	almost unlimited
GLM ASR	✅	100
MiMo	✅	supports context injection
ElevenLabs	❌	-
Qwen3-ASR	✅	suggested limit: 100
FunASR Nano	✅	suggested limit: 100
Other local models	❌/depends on engine	depends on engine capability

Local hotwords

When using Qwen3-ASR or FunASR Nano, changing hotwords may require the local model to be prepared again before the new list fully takes effect.

Continuous Speaking Mode Pro

A continuous dictation mode with automatic start/stop: after activation, recording starts when speech is detected and stops after silence, no repeated tapping needed.

Highlights

• Auto start: recording starts automatically when speech is detected after activation
• Auto stop: stops when silence is detected and runs recognition
• Continuous: multiple rounds without tapping repeatedly
• Follows auto-stop settings: silence window and sensitivity follow ASR Settings → Auto-stop on Silence; no separate parameters

How to use

1. Open the BiBi Keyboard Pro panel
2. Add the continuous mode button under Settings → UI Settings → Keyboard UI → Custom keyboard layout, then toggle it on the keyboard
3. Activate the mode via the button and start speaking; recording starts automatically on speech and stops after silence
4. Tune the silence window (default 1200 ms) and sensitivity (default 4) under Settings → Smart → ASR Settings → Auto-stop on Silence

The floating ball also supports continuous speaking mode, which is useful for dictating multiple segments in the current app or while using another IME. It keeps listening locally and submits each segment after silence. And you can toggle the switch in the floating ball menu.

Notes

• Continuous mode keeps the mic listening longer and increases battery usage.
• In very noisy environments it may false-trigger; consider switching back to press-and-hold.
• It may take a word or two to detect speech and start recording; try to use locally captured pre-trigger audio to reduce dropped beginnings.

Input Field Context Pro

When using AI post-processing from the main keyboard, Pro can send text around the cursor as reference to the LLM. This helps the model understand continuity, terminology, and tone. With the IME Hook module enabled, floating-ball recordings can use the current input-field context too. See IME Bridge Module for setup and privacy boundaries.

How to use

1. Open Settings → Smart → AI Feature Settings
2. Enable Use input field context (Pro)
3. Dictate from the main keyboard with AI post-processing enabled. To use it from the floating ball, also enable Settings → Input → More Input Methods → Advanced → IME Hook module

Privacy

This sends nearby input-field text to the selected LLM provider as reference. Enable it only when you trust that provider and the current content is appropriate to send.

Scope

Input field context applies to AI post-processing after main-keyboard dictation and to floating-ball dictation when IME Bridge is enabled. MiMo multimodal ASR can also use Pro context information to improve proper nouns and scene-specific terms.

Offline Traditional Chinese Conversion Pro

Automatically converts Simplified Chinese transcripts to Traditional Chinese.

Highlights

• Smart conversion: offline conversion algorithm
• Seamless: runs automatically without hurting recognition speed
• Global: works for all providers and input scenarios

How to use

1. Open BiBi Keyboard Pro settings
2. Go to Settings → Input → Input Settings → Pro features → Convert recognition result to Traditional
3. Choose a rule:
   • Standard Traditional
   • Taiwan usage
   • Hong Kong usage

Note

Conversion applies after all other text processing (including AI post-processing), ensuring final output is Traditional Chinese.

Regex Post-processing Pro

Apply regex rules to recognition results for advanced text transformations.

Highlights

• Full Java regex support
• Chained rules: apply multiple rules in order
• Templates: one-tap common templates with auto-filled notes
• Notes: notes are for display/management only (do not affect matching)

How to use

1. Open settings
2. Go to Settings → Smart → ASR Settings → Result Optimization (Pro)
3. Enable and tap Add rule (or use Templates)
4. Fill in:
   • Pattern: regex
   • Replacement: replacement text ($1, $2, etc.)
   • Options: ignore case, multiline, etc.
   • Note: a label for list display (does not affect behavior)
5. Save; rules apply sequentially

Note

By default, regex post-processing runs after AI post-processing. If you enable "Run regex before AI post-processing", the regex-processed text becomes the AI input, which is useful when you want to clean fixed noise before asking AI to rewrite it.

App-specific Prompt Pro

This feature is only available in Pro

Assign different AI post-processing prompt presets per app. Example: use "General post-process" for chat apps, "Basic polishing" for notes, and "Extract to-dos" for task tools.

How to use

1. Open Settings → Smart → AI Feature Settings
2. Tap "App-specific Prompt (Pro)"
3. Enable "App-specific Prompt"
4. Tap "Add app" and select a prompt preset for that app

Notes

This requires enabling BiBi Keyboard’s accessibility service to detect the foreground app. It does not read your chat content or screen text.

AI Assistant Pro

Trigger voice commands with a wake word, so AI can directly handle translation, writing, editing, summarization, and more.

Highlights

• Wake-word trigger: say the wake word at the beginning of speech to enter AI Assistant flow
• Preset modes: enable multiple presets at once (for example translation/writing/editing), then match by keywords
• Fuzzy matching: wake words and preset keywords support fuzzy matching to reduce trigger failures caused by minor slips
• Customizable: customize wake words, keywords, and prompts

How to use

1. Open Settings → Smart → AI Feature Settings
2. Enable AI Assistant (Pro)
3. Configure wake words and preset keywords
4. Speak in the form of "wake word + command" to trigger (for example: "Dian Dian, translate this into English")

Advanced UI Theming Pro

On top of Material 3 dynamic color, Pro adds more color customization options.

What’s included

OSS:

• Material 3 dynamic colors

Pro adds:

• Built-in theme colors: pick theme colors independent of system wallpaper
• AMOLED black mode: use a deeper black background in dark mode
• Monet toggle: choose whether to follow wallpaper dynamic colors
• Key color customization: choose default, built-in, or custom colors as the theme seed

How to use

1. Open settings
2. Go to Settings → UI Settings → App UI → Theme mode (Pro)
3. Configure:
   • Appearance: system / light / dark / AMOLED black
   • Monet: whether to use wallpaper dynamic color
   • Key color: default, built-in color, or custom color
   • Color style / spec: adjust the generated theme palette

Note

Choosing a custom key color overrides pure Material 3 dynamic color behavior. Restore the default key color to return to the default theme behavior.

Omni-direction Cursor Sliding Pro

Enables omni-direction cursor movement via sliding on the space bar.

How to use

1. Enable Settings → Input → Input Settings → Pro features → Cursor Glide Mode
2. On the keyboard, start from the space key area and swipe up to trigger cursor sliding
3. The mic icon turns into four directional arrows
4. Swipe up/down/left/right to move the cursor; the cursor follows your finger position

Swipe up from the space key to enter the mode. It is also available from the AI edit panel.

WebDAV Auto Backup Pro

Automatically backs up all settings to WebDAV storage for multi-device sync. Enable it under Settings → System → Backup and Sync → Auto Backup (Pro); the backup interval is an interval value plus a unit (hours/days), minimum 6 hours. The page shows the next scheduled backup time and the last backup status.

See Backup and Sync for configuration details and troubleshooting.

Get Pro

Pro is distributed via Google Play. See Activation for purchasing, and OSS vs Pro Comparison for feature differences.