After Your First Dictation¶
You have already completed the essential OpenScreech workflow. This page explains the controls behind it and helps you choose what to learn next.
How Hotkeys and Profiles Connect¶
Hotkey → mapped Profile → Processing → Delivery
A hotkey defines:
- Which key activates dictation.
- Whether activation is push-and-hold or toggle.
- Which profile starts the session.
A profile defines:
- Word replacement and optional LLM processing.
- Auto Tag and trigger routing.
- Optional Screechback audio responses.
- Automatic paste and paste formatting.
- Temporary history.
- Saved Outputs and Notes.
- Overlay position, colours, meter behavior, and quick-switch hints.
Selecting a profile in Manage Profiles... does not remap your hotkeys. Open Hotkeys -> Manage Hotkeys... to see or change the profile attached to each hotkey.
Choose an Activation Style¶
| Style | How It Works | Good For |
|---|---|---|
| Push and Hold | Hold the key while speaking; release it to finish | Short dictation, quick prompts, and precise capture boundaries |
| Toggle | Press once to begin and again to finish | Longer speech without holding a key |
New setups seed Right Command as push-and-hold and Right Option as toggle. They initially point to the same profile, so the activation style changes without changing the processing behavior.
Read and Move the Overlay¶
The overlay's status line tells you what is happening:
| State | Meaning |
|---|---|
Checking microphone |
Preparing the audio input |
Preparing speech model… |
Recording has started and audio is being retained while the local speech model finishes loading |
Listening (Profile Name) |
Recording with the named profile |
Transcribing |
Converting captured speech to text |
Processing with ... |
An enabled LLM post-processor is running |
Pasting |
Inserting the finished result into the active application |
Ready (auto paste disabled) |
Text is finished, but the profile is not set to paste it |
The transcript area follows the newest preview text, and the audio meter responds to microphone level. Profiles can optionally show a peak marker. The meter can move during Preparing speech model…; this confirms capture is active even though preview transcription has not started yet.
Drag the live overlay wherever it is comfortable. If Save dragged overlay position is enabled on the active profile, OpenScreech remembers that relative screen position. Turn it off to use the profile's Overlay position choice again.
Recall Recent Results¶
When the active profile has Add transcriptions to temporary history enabled, OpenScreech keeps up to ten finalized results in memory.
To reuse one:
- Start dictation with the same hotkey.
- Press
Left Arrowto replace the live capture with the newest history result. - Continue pressing Left Arrow to move farther back.
- Press
Right Arrowto move forward. - Finish with the hotkey's normal action to paste the displayed result.
The overlay shows Recalling · 1 back, Recalling · 2 back, and so on.
Important behavior:
- Arrow keys select a result but do not paste it.
- Recalled text is already finalized and is not transcribed or processed again.
- Entering recall discards the unfinished live capture.
- Pressing Right Arrow beyond the newest result returns to fresh listening.
Escapecancels without pasting.- Temporary history is cleared when OpenScreech exits.
Persistent Notes are separate from temporary history.
Check Your First Profile¶
Now that basic dictation works:
- Open
Profiles->Manage Profiles.... - Select the profile shown in your overlay.
- Locate
Auto paste into active windowandAdd transcriptions to temporary history. - Keep advanced processing off until you deliberately choose to test it.
- Return to
Manage Hotkeys...and confirm which hotkeys map to this profile.
This gives you a known working baseline before you create specialized profiles.
Choose What to Do Next¶
Speak directly into AI tools¶
Use Dictate Prompts for AI Tools for Codex, Claude Code, editors, terminals, and similar tools.
Improve names, punctuation, or formatting¶
- Use Word Replacement Libraries for exact, predictable substitutions.
- Use LLM Post-Processing for cleanup, restructuring, extraction, and model-assisted transformations.
Keep a searchable voice collection¶
Use Notes to retain selected transcripts, tags, and audio beyond the temporary history.
Hear a summary or response¶
Use Screechback to play a final result or one LLM-stage result through a downloaded local speech model. The Hear Summaries and Responses preset is a practical first setup.
Record a meeting or long session¶
Use Long-Form Recording for durable audio, transcript, subtitle, timing, and optional speaker-labelled artifacts.
Transcribe existing media¶
Use Transcribe File(s) for queued audio and video transcription.
Run voice-controlled actions¶
Use Trigger Routing Libraries only after ordinary dictation is predictable and tested.
Deeper Reference¶
Quick Tips¶
- While OpenScreech is listening, press another profile's quick-switch key to change the final processing and delivery. This is useful when you start in a general mode and realize the result should be formatted, saved, or routed differently.
- Overlay placement and selected-text capture begin with the original profile, so switching profiles does not restart those capture-time behaviors.
Visual Notes¶
- TODO: add a GIF of Left and Right Arrow history recall.
- TODO: add a screenshot comparing push-and-hold and toggle hotkeys.
- TODO: add a screenshot showing the overlay position controls.