Dictation with your own API key
Already pay for OpenAI, Groq or another AI provider? Add your API key to OiPer and use their models for transcription, for formatting, or both. Your audio goes straight to the provider you chose, they bill you directly, and the key never reaches OiPer.
Updated
Use the providers you already have
Register a speech model, a language model, or one of each.
Pre-filled providers
OpenAI, Anthropic, OpenRouter, Hugging Face, Groq, Cerebras, Fireworks, Kilo and OpenCode come with their base URL and usual environment variable filled in.
Read about ModelsAny OpenAI-compatible endpoint
Enter the base URL yourself to use any other OpenAI-compatible service, including a server running on your own machine.
Read about ModelsTranscription and formatting
Speech models (ASR) transcribe your voice. Language models (LLM) clean up, rewrite or translate the transcript before it is typed.
Read about FormattingYour dictionary applies
Custom words are passed to the speech model in cloud mode, so names and jargon are more likely to come out right.
Read about DictionaryMix local and cloud
Each profile has its own hotkey and its own mode.
Local speech, cloud formatting
Transcribe on your computer with local Whisper and send only the transcript to your language model for formatting.
Read about ProfilesA cloud hotkey next to a local one
Keep an offline local profile on one shortcut and a cloud profile with your key on another, and pick per dictation.
Read about ProfilesFully local formatting
Point the formatting model at an OpenAI-compatible server on your own machine so neither audio nor text leaves it.
Read about FormattingHow your keys are stored
Kept apart from your settings, and sent only where they belong.
Separate from your config
Keys are stripped from config.json and written to auth.json in the same folder, so you can share or sync your settings without leaking them.
Read about Security & Data HandlingOr use an environment variable
auth.json is protected by your user account’s file permissions, not the system keychain. Point an entry at an environment variable and OiPer never stores the key at all.
Read about Security & Data HandlingSent only to your provider
A key is sent to the provider it belongs to, as part of a request you triggered. It is never transmitted to OiPer.
Read about Security & Data HandlingSet it up
Three steps on the Models and Profiles pages.
1. Add a model
On the Models page, add a speech or language model with its provider, base URL, model name and API key. For Groq, for example, the base URL is https://api.groq.com/openai/v1 and a model is whisper-large-v3-turbo.
Read about Models2. Switch a profile to cloud
Set the profile’s mode to cloud and choose your speech model, or keep it local and pick your language model for formatting.
Read about Profiles3. Dictate
Hold the profile’s hotkey and speak. If a request fails, check that the base URL and model name match what your provider expects.
Read about TroubleshootingQuestions, answered
What people ask before they start dictating with OiPer.
Can I use my own OpenAI API key for dictation?
Yes. Add an OpenAI speech model with your key on the Models page and set a profile to cloud mode. OpenAI bills you directly for the usage.
Can I use Groq Whisper for dictation?
Yes. Groq is a pre-filled provider in OiPer. Add your Groq key and a Whisper model such as whisper-large-v3-turbo, and dictation runs on Groq.
Do I need a paid OiPer plan to use my own key?
No. Paid plans are for OiPer’s hosted models. With your own key you pay only your provider.
Does OiPer see my API key?
No. Keys are stored on your computer in auth.json, or read from an environment variable, and are sent only to the provider they belong to.
Can I use a local LLM for formatting?
Yes, if it exposes an OpenAI-compatible endpoint. Enter its base URL and model name on the Models page and choose it for formatting.
What is sent to my provider?
In cloud mode, the recording goes to your speech provider, and the transcript goes to your language model when formatting is on. Nothing is sent in the background.
Can I use OpenRouter or Anthropic?
Yes, for formatting. Both are pre-filled providers you can register as the language model that cleans up, rewrites or translates your transcript.