In-App Help Cues (? Buttons and Hover Tooltips)¶
Why This Matters¶
OpenScreech includes built-in guidance directly in the UI.
If you are unsure what a setting means, check these first:
?question-mark buttons beside labels- hover tooltips on many toggles/buttons/fields
These cues are designed to explain behavior at the point of use.
OpenScreech Home also provides hover labels for destinations in its compact icon sidebar and for its sidebar, pop-out, dock-back, and bring-window-forward controls.
How The ? Help Works¶
In OpenScreech, ? help supports two interaction styles:
- Hover over
?to see quick help text. - Click
?to open a popover with the same explanation.
This is useful for both quick scanning (hover) and focused reading (click).
Where You Will See It¶
? and hover help are present in major configuration windows:
App SettingsRecording SettingsHTTP Transcription ServiceManage Profiles...Manage Hotkeys...Word Replacement Libraries...LLM Post-Processing Library...Trigger Routing Libraries...Transcribe File(s)...
High-Value Help Topics You Should Look For¶
These are especially helpful for first-time users:
Profiles¶
- Quick switch key behavior (switches profile during active dictation; does not start/stop)
- Overlay position vs
Save dragged overlay position Paste input formatvsPaste format- Temporary history retention and Left/Right Arrow overlay recall
- Screechback source, speech-model, voice, and language selection
- PocketTTS 6-layer/24-layer tradeoffs and voice catalogue metadata
- Saved output minimum duration threshold
LLM Post-Processing Library¶
- Provider differences (
OllamavsOpenAI Compatible) - Endpoint, API key, OpenAI-compatible custom headers, model, timeout, optional system-prompt delivery, and wrapping behavior
- Shared-processor effects across profiles, Meeting Recording, and Transcribe File(s)
Hotkeys¶
Push and HoldvsToggleactivation type- Activation key mapping behavior
- Hotkey enable/disable meaning
Word Replacement Libraries¶
- Rule execution timing (before LLM)
- Match semantics and case sensitivity
- Surrounding whitespace/character cleanup behavior
Trigger Routing Libraries¶
- Split routing behavior
- Independent route evaluation behavior
- Multi-trigger chunking behavior
HTTP Transcription Service¶
- Bind address implications (
127.0.0.1vs0.0.0.0) - Token requirement behavior
- Save request logging behavior
- Endpoint path override behavior
Recording Settings and Transcribe File(s)¶
- Output format meanings (
TXT,SRT,WebVTT,TTML,SMPTE-TT) - Reusable LLM post-processor selection
- Single Meeting processor input selection and saved-output requirements
- LLM output and payload capture behavior
- Cue split options and chunking semantics
- Speaker diariser selection and explicit model-download behavior
Local Speech Models and Voices¶
- The
Local Speech Models & VoicesApp Settings heading explains that these on-device text-to-speech resources are distinct from speech recognition and LLM post-processing models. - Speech-model vs large-language-model terminology
- Download contents, upstream source, and local destination
- Kokoro, PocketTTS, and Supertonic-3 voice guidance
- First-test model-loading delay
Show in Finderfor installed resources
App Presence¶
Show OpenScreech in Dock: makes OpenScreech available in the Dock and Command-Tab.- Clicking the Dock icon brings existing windows forward or opens OpenScreech Home when no regular window is visible.
- Turning the setting off preserves the menu-bar interface.
Exact In-App Message Examples¶
These are examples of real help wording shown in the app, so you can match docs to UI text quickly.
Profiles Window Examples¶
Quick switch key: "Switches profile while dictation is already active. It does not start or stop dictation."Transcribed Text Wrapping: "Adds tags around transcript text before sending it to the LLM."Paste format: "Auto (Multi-format)writes plain text + rich text + HTML and the target app chooses the best supported format."Add transcriptions to temporary history: "Temporary history keeps the last 10 finished transcriptions in memory only. Use the History menu to copy an item, or during dictation press Left/Right Arrow to browse paste-ready results in the overlay. Arrow keys only select; finish with the active push-and-hold or toggle hotkey to paste the displayed result. Recalled results are not processed again."
Recording Settings Examples¶
Processor Input: "Choose the single recording representation sent to the LLM processor as its user message. Original Transcript is always available. Other inputs are available only when their matching output is enabled and saved for the recording."Word-Level Timing Sidecar (.words.jsonl): "Writes one JSON object per recognized word, including its chunk, start and end times, and confidence. Requires live transcription and can optionally be used as the LLM processor input."
Hotkeys Window Examples¶
Activation Type: "Push and Hold: hold to dictate. Toggle: press once to start, again to stop."Enable Hotkey: "Disables runtime handling for this hotkey without deleting its configuration."
Word Replacement Libraries Examples¶
Word Replacements: rules run before LLM and optional cleanup can remove whitespace/characters around matches.Remove Chars: removes only contiguous listed characters around a match.
Trigger Routing Libraries Examples¶
Split Routing: enables splitting one dictation into multiple routing chunks when multiple triggers match.Trigger Routes: routes are evaluated independently and all enabled matches run; with split routing, phrase triggers create chunks andAny Textcan run per resulting chunk.
HTTP Service Settings Examples¶
Bind Address: localhost for local-only access, all interfaces for LAN exposure.Require Token: clients must sendAuthorization: Bearer <token>when enabled.Save Folder: request artifacts are stored in timestamped subfolders.
Recording Settings / Transcribe File(s) Examples¶
Processor: choose the reusable LLM post-processor used to create the.llm.txtoutput; its configuration is managed in the LLM Post-Processing Library.Cue split: controls how transcript text is broken into subtitle cues.Chunking: segments transcript output windows; ASR still handles internal streaming/chunking.Save LLM Request Payload: saves the exact JSON request body OpenScreech sends using the selected LLM post-processor; it can contain transcript and rendered prompt content.Diariser: selecting a FluidAudio model does not download it; use the selected model's Download button and wait forReady.
App Presence Example¶
Show OpenScreech in Dock: "Makes OpenScreech available in the Dock and Command-Tab. Clicking its Dock icon opens OpenScreech Home when no other OpenScreech window is visible."
Suggested Learning Habit¶
When configuring a setting you do not recognize:
- Hover the label/control first.
- Click the nearby
?button if present. - Apply one change at a time.
- Test with a short dictation sample.
This reduces confusion and prevents stacking multiple unknown changes at once.
Quick Tips¶
- Hover help is best for a quick reminder; clicking
?keeps the same explanation open while you compare it with nearby controls. - Test after one meaningful setting change. Several individually valid changes can interact, making the original cause harder to identify.