I use dictation daily for writing and prompting agents. Existing dictation apps worked well for me, but recurring cost and unclear data handling pushed me to build my own: Utter

Core idea

  - Free with local models

  - Also free with BYOK

  - Sub-second transcription and post-processing

  - More than transcription: custom prompt-based post-processing

What it does

  - Dictation with customizable shortcuts, including toggle or hold-to-talk

  - Modes let you save different model + prompt setups for different workflows

  - Last-used mode remembered per app & per-mode shortcuts

  - Meeting recording with speaker-labeled transcripts, summaries, and action items

  - Audio/video file transcription, watched folders

  - Local audio/transcript retention & auto-export of transcripts

  - Context-aware post-processing, with visibility into exactly what context was used

  - Link summarization (Web, YouTube, PDF, DOCX, PPT) + follow-up AI chat

  - Markdown note editor

  - iPhone app with dictation keyboard

Privacy, custom post-processing, and frictionless phone-to-desktop capture were major motivators for me: I wanted to record ideas on walks, have longer recordings turned into clean Markdown notes, and avoid sending personal voice notes to the cloud. Utter requires no account or registration, syncs through iCloud, supports fully local workflows (for example, local transcription with Parakeet plus local post-processing with Ollama or LM Studio), and includes a built-in Markdown editor.

Curious to hear what dictation app users think, especially where this still falls short.