More context for your AI. Less time at the keyboard.
The quality of an AI response often depends on the context in the prompt. Voice makes it easier to explain the background, desired outcome, constraints, examples, and acceptance criteria without reducing the request to a few rushed sentences.
Why dictate an AI prompt?
Typing creates a subtle pressure to be brief. That can be helpful, but it can also remove the facts an AI assistant needs to make a good decision. Speaking lets you describe the situation at conversational speed, closer to the way you would brief a colleague.
Elokin transcribes that explanation locally, then lets you keep the raw version or process it into clearer written text. You review the result before pasting it into the destination service.
A repeatable prompt structure
- Explain the current context and what already exists.
- Describe the outcome, constraints, and what must not change.
- Specify the expected format and how success should be verified.
Pausing briefly between those parts gives the processing profile a natural structure to work with. You do not need to speak punctuation commands when the profile can infer ordinary sentence and paragraph boundaries.
Example outline: “Here is the context… The problem is… I want the result to… Keep these constraints… Finally, verify it by…”
Choose the right profile
Raw keeps the transcript closest to your speech. It is useful when domain detail matters more than style. Polished improves punctuation and organization without deliberately shortening the request. Concise is useful for removing repetition after you have explored an idea aloud.
Custom profiles can apply a prompt template or format the dictation for a particular AI tool. A local profile keeps processing on the Mac; a cloud profile sends the transcript to the provider you explicitly configured.
Prompting without wasting transcription tokens
A cloud language model does not need to convert the original audio. Elokin performs speech recognition locally, so the AI service only receives the text you decide to submit. That separates transcription from reasoning and avoids using provider capacity for a task your Mac can perform.
The submitted text will still count toward the destination service’s token usage. The saving is specifically on speech recognition and optional local cleanup—not on the final prompt sent to ChatGPT, Claude, Codex, or another model.
Frequently asked questions
Can I use Elokin with any AI assistant?+
Yes. Elokin copies ordinary text that you can paste into desktop apps, browser tools, coding assistants, or chat interfaces.
Does local dictation consume AI tokens?+
No. Local speech recognition does not call an external language model. Tokens are used only when you submit the finished prompt to an AI service or select a custom cloud profile.
Should I use Raw, Polished, or Concise?+
Raw is useful when every spoken detail matters. Polished usually works well for readable prompts. Concise is useful when repetition would distract from the request.