I built utter, a macOS + iOS dictation app that focuses less on raw transcription and more on what happens after dictation.

Standard dictation is fast, but the output usually needs cleanup. Utter lets you define custom prompts that transform your spoken input before it’s inserted — e.g. removing filler, adding structure, formatting as bullet points, emails, or notes.

It works system-wide: press a hotkey, speak, and the processed text is inserted at the cursor in any app (Notes, editors, terminals, browsers, etc.).

A few details:

- Works out of the box with no setup using built-in cloud models

- Optional local models or BYOK for those who prefer that path

- Transcripts can be saved locally in formats like Markdown

- iOS companion app with iCloud sync

- No accounts, no login, no data retention

I built this primarily for my own workflows (notes, writing, coding prompts) and would love feedback on what’s missing or where it breaks down. One of my favorite uses is dictating prompts for agentic coding workflows, which are then post-processed using a custom prompt that includes a repository file map for semantic search and automatic @file tagging.

Link: https://utter.to

Happy to answer questions.