Text Enhancements settings

Open Settings → App → Text Enhancements to control everything Echo applies to text around the raw transcription: local cleanup, on-screen context, correction rules, vocabulary hints, and system spelling.

Automatic cleanup

BudgieScribe cleanup enables Echo's local post-processing engine, which fixes punctuation, capitalization, and common recognition mistakes after transcription. The status line next to the toggle reports whether the engine is running, installed but waiting to start, checking, or lacking a cleanup language.

Cleanup work happens in cleanup languages you download on this device. Each installed language is a local model; installing more languages covers more of the languages you dictate in, and transcription in a language without a cleanup model passes through unpolished. The onboarding wizard asks for at least one cleanup language; this panel manages the list afterwards.

Transcription context

Read the active window lets Echo use what is on screen when a dictation starts to get names and specialist terms right. When off, nothing is read — not even the application name. When on, Echo reads the accessibility tree of the window where dictation starts; that capture uses the CPU and system APIs, and nothing leaves your machine. Sandbox store builds cannot read window text and use only the application name and window title.

Replacements

Replacements are correction rules applied to transcription results, with an optional scope:

• a source pattern to match, where * acts as a wildcard;
• the replacement text — an empty replacement deletes the matched text;
• where it applies: Everywhere, One application (by application name), or One activity (such as email, chat, meeting, docs, or code);
• whether the match is Case sensitive;
• whether the match must be a Whole word.

Search the list, add entries, edit or disable them without deleting, and delete the ones you no longer need. Replacement rules apply after transcription and can also apply to workflow results, so a broad pattern can change text you intended to keep. Run a small test recording after every change.

Vocabulary

Vocabulary lists are words and phrases hinted to the speech model during transcription — names, product terms, jargon. Organize them in lists, enable the list you need (inactive lists stay configured but unused), and add or remove entries at any time. Words in your vocabulary are never touched by replacements or spelling correction.

System spelling correction

Use the system spell checker applies the spelling corrections your operating system already provides to transcription results. Echo probes which dictionary languages are installed and shows them as badges; spelling only — there is no public interface to grammar engines. The toggle is disabled with an explanation when the system offers no usable spell checker or no language is installed.

Where these features run

Cleanup, replacements, vocabulary, and spelling all run locally. Before pointing them at sensitive text, review Privacy & Data for storage details, and use Diagnostics when sharing configuration examples.