Epic: match the paid dictation apps, running on your own computer

Open
  • Is something wrong?
#166 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
5/5
Estimated time
Over a week
Newbie friendliness
8/100
Issue type
Feature
Clarity
Needs clarification
Activity status
Active
Tech stack
ollama, swift
Domain
ai, desktop

Research direction

This is a roadmap epic with no code location named, so there is no single starting file. Read the linked child issues, beginning with #118, the local AI layer that the tone, command mode and meeting features depend on, and pick one scoped child issue to work on. Done means that child issue's own acceptance criteria are met, not this epic being closed.

Written by the indexing model from the issue text.

Description

enhancement priority:P1 roadmap

Goal: everything the paid dictation apps charge a monthly subscription for, free, with the work happening on your own machine instead of someone else's servers.

They're two halves of a product. We have most of the first and none of the second.

Dictation

What the paid apps do Where we are
Punctuation, filler removal, tidy sentences Done
Personal dictionary for names and jargon Done, manual entries
Dictionary that learns when you correct a word #151
Snippets: say a phrase, get your address or boilerplate #167
Tone that fits the app you're writing in #168
Command mode: select text, speak an edit #169
Works in every app, no account Done
Offline Done, and they can't

Meeting notes

The paid ones now listen to system audio without joining a call, write a summary with action items, label who said what, and answer questions about past meetings. Nothing of ours does this yet:

  • #170 capture system audio on each system
  • #171 transcribe a long meeting while it happens
  • #172 summary, action items, and questions about a meeting
  • #80 who said what

The piece everything else needs first

#118, a local AI layer: talk to Ollama or any compatible endpoint, let people pick a model, and fail back to the plain transcript whenever it doesn't answer. Tone, command mode, summaries and meeting questions are all prompts on top of that one piece.

Order

  1. The AI layer (#118), then tone (#168) and command mode (#169), which are small once it exists.
  2. Snippets (#167) and the learning dictionary (#151), which need no AI at all.
  3. Meeting notes, in order: capture (#170), long transcription (#171), then summaries and questions (#172).

Meeting notes are the biggest lift in the project's history: audio capture works differently on all three systems, and an hour-long meeting is a different problem from a twenty-second dictation. Worth being honest about that rather than promising it next month.

What we won't copy

Their free tiers cap how many meetings a week you get and keep your transcripts on their servers. Ours has no caps, and nothing leaves your computer.

Dominant language
Swift
Stars
540
Forks
74
No merged PRs in 30d

Getting set up

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from karansinghgit/speaktype

All issues in karansinghgit/speaktype

Similar issues

More Swift issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.