AI Assistant

AI Assistant listens locally and suggests likely lyrics and Bible passages in Suggestions, beside Live View. You choose what goes live. Follow automatically additionally lets AI navigate within the current section’s permitted sources.

Set it up before rehearsal

1. Open AI Assistant in the title bar and choose Set up & test.
2. Under Computer setup, use the recommended setup for this computer. Apple Silicon uses Metal; other available setups are labelled in the app when their support is experimental.
3. Open Required downloads and finish every required model download.
4. Under Languages, configure Dutch-capable speech recognition if Dutch will be spoken or sung. Understanding a Dutch reference does not require a Dutch Bible; it does require a multilingual speech model.
5. Use the title-bar Microphone control to choose and test the input. Restart AI Assistant after changing the device.
6. Open Suggestions beside Live View. Configure a section’s AI mode and matching scope when you want saved following behavior or a restricted search.
7. Turn on AI Assistant and test the actual microphone, room, and service content before going live.

Suggestions work on manually operated cues too. A section or cue explicitly set to Off stays excluded. Existing saved AI modes can allow following; turning the Suggestions switch off overrides those saved choices until you explicitly enable it again.

Choose the right AI mode

For either song-finding mode, Match lyrics against limits detection to songs in the section or expands it to the full lyrics database. Use the narrowest mode and matching scope that fit the service.

Operate it live

• Open Suggestions for AI matches and Worship Assist requests. Their count remains visible on the tab; Controls retains cue, slide and screen buttons.
• Show adds or reuses an ordinary cue in the live section and starts presenting it. Bible cues use the same Bible system as Cmd+K / Ctrl+K, including exact text from the selected translation. Show and follow also enables following. The × dismisses a suggestion.
• The wide Follow automatically switch turns automatic actions on or off. Turning it off keeps suggestions available and remains in effect across cue changes.
• Manual navigation while following temporarily holds automatic actions. The AI pill shows when following resumes, with a shrinking progress bar. Suggestions continue during this hold.
• The default hold is 5 seconds, adjustable from 5 to 120 seconds in Settings → AI Assistant → Manual control. Switching following off cancels automatic resumption.
• Pause AI inside the status pill stops listening and automatic actions indefinitely. Resume returns to suggestions. Worship Assist retains its own acceptance preferences.
• The microphone opens during presentation while AI is enabled, including manual cues. Explicitly excluded sections/cues and Pause stop capture. Watch performance warnings and use a lighter speech model if audio falls behind.

If AI does not move the deck

Check these in order:

1. The title-bar AI status says it is on and ready.
2. The Suggestions follow switch is on and the section allows the detected source. Section-only matching never searches outside its prepared cues. Unplanned songs retain their existing automatic-add permission; unplanned Bible following has its own section setting.
3. The correct microphone is selected, its level moves during speech or singing, and operating-system microphone permission is allowed.
4. Every required model download is complete.
5. The chosen language matches the content being spoken or sung.
6. AI is not paused, explicitly excluded or still in a manual-control hold.

If recognition becomes slow or audio queues up, open AI Assistant → Computer setup and use the recommended or lighter setup. Use Settings → AI Assistant → Troubleshooting only when you need the live debug panel.

Recognize Bible passages

Spoken references such as “John chapter three verses sixteen through twenty” and “Johannes 3 vers 16 tot en met 20” resolve to the same passage in your default Bible. No multilingual passage model is required for named references. Install the desired Bible through the existing Bible library; if the default is unavailable offline, the suggestion lets you choose an installed translation.

Wording recognition searches the local song library and installed offline Bibles, including imported USFM text. Under Settings → AI Assistant → Passage recognition, download the optional multilingual recognition model (135 MB) for quotations whose wording differs or is spoken in another language. Presenter downloads prepared recognition indexes for installed KJV, ASV and WEB Bibles and keeps them on this computer for offline reuse. It prepares custom songs, imported Bibles and changed passages locally. If the prepared download is unavailable, local preparation remains available and the status explains the download failure. Local calculation pauses during live microphone capture; cached passages remain searchable. Let preparation finish before the service. You can cancel the download or remove the model from the same disclosure.

Possible paraphrases always require your confirmation, even while following. Brief generic phrases and uncertain matches may produce no suggestion. Presenter projects stored Bible text, never AI-written verses. Test your speech model and translations during rehearsal; recognition cannot be guaranteed for every quotation.

Privacy

Recognition audio is processed locally. Presenter does not send microphone audio to a cloud transcription service and does not write spoken transcript text to routine logs.