Custom OpenAI-Compatible Endpoints

HyperWhisper can send your transcripts to any server that supports the OpenAI chat-completions API. That server does the AI post-processing. You can use a local model on your own machine, a self-hosted server, or a hosted gateway such as OpenRouter. After you configure the endpoint, your Modes need no changes.

What custom endpoints do

A custom endpoint is an alternative post-processing provider. A Mode can have AI cleanup on, with a custom endpoint as its provider. HyperWhisper then sends the raw transcript to your server in the standard OpenAI chat-completions format. HyperWhisper uses the reply as the cleaned-up output. Custom endpoints are not part of the transcription. They do only the AI cleanup and formatting pass that comes after speech recognition.

Adding an endpoint

macOS
Windows

Open the Model Library

Open HyperWhisper. Then click Model Library in the sidebar.

Find the OpenAI-compatible endpoints card

Scroll to the OpenAI-compatible endpoints card near the bottom of the page. Click Add endpoint.

Choose a provider tab

The sheet has three tabs:

LMStudio — fills in the base URL http://localhost:1234/v1 and the name “LMStudio”. This tab finds your running models automatically.
Ollama — fills in http://localhost:11434 and the name “Ollama”. This tab finds your running models automatically.
Custom — for OpenRouter, other hosted APIs, or any other OpenAI-compatible server. You enter the base URL and the model name manually.

Fill in the fields

Field	What to enter
Name	A label that you recognize, for example “My Ollama Server” or “OpenRouter GPT-4o”
Base URL	The base URL of your server (read URL format)
Model	The model identifier that your server expects, for example `llama3.2` or `mistral-7b-instruct`
API Key	Optional. If your local server does not need authentication, leave this field blank.

On the LMStudio and Ollama tabs, HyperWhisper reads the models list from the server. If the models load, select one from the dropdown. If they do not load, type the model name.

Test the connection (recommended)

Click Test connection before you save. HyperWhisper sends a small request to your server and shows the response. A green indicator means that the server answered correctly.

Save

Click Add Endpoint. If you edit an endpoint, click Save Changes instead. The endpoint appears in the card list.

Open the Model Library

Open HyperWhisper. Then click Model Library in the sidebar.

Add a custom endpoint

Click Add endpoint in the Custom Endpoints section.

Choose a provider tab

The window has three tabs:

LMStudio — fills in the base URL http://localhost:1234/v1 and the name “LMStudio”. This tab finds your running models automatically.
Ollama — fills in http://localhost:11434 and the name “Ollama”. This tab finds your running models automatically.
Custom — for OpenRouter, other hosted APIs, or any other OpenAI-compatible server. You enter the base URL and the model name manually.

Fill in the fields

Field	What to enter
Name	A label that you recognize, for example “My Ollama Server” or “OpenRouter GPT-4o”
Base URL	The base URL of your server (read URL format)
Model	The model identifier that your server expects, for example `llama3.2` or `mistral-7b-instruct`
API Key	Optional. If your local server does not need authentication, leave this field blank.

On the LMStudio and Ollama tabs, HyperWhisper reads the models list from the server. If the models load, select one from the dropdown. If they do not load, type the model name.

Test the connection (recommended)

Click Test connection before you save. HyperWhisper sends a small request to your server and shows the response. A green indicator means that the server answered correctly.

Save

Click Add Endpoint. If you edit an endpoint, click Save Changes instead. The endpoint appears in the model list.

URL format

Enter the base URL of your server. On the Ollama and LMStudio tabs, HyperWhisper adds the correct path for you. For the Custom tab, read the platform notes after this table:

Provider tab	Path appended	Example full URL stored
Ollama	`/v1/chat/completions`	`http://localhost:11434/v1/chat/completions`
LMStudio	`/chat/completions`	`http://localhost:1234/v1/chat/completions`
Custom (Windows)	`/chat/completions` (only when the URL does not end with it)	`https://openrouter.ai/api/v1/chat/completions`
Custom (macOS)	(none — HyperWhisper uses your URL exactly)	`https://openrouter.ai/api/v1/chat/completions`

On Windows, the Custom tab adds /chat/completions when your URL does not end with it. On macOS, HyperWhisper stores your URL exactly as you enter it. If your server needs the full path, enter the full path.

API key storage

HyperWhisper stores API keys securely. It never stores them in plain text:

macOS — HyperWhisper stores the key in the system Keychain.
Windows — HyperWhisper stores the key in Windows Credential Manager.

To change or delete a key, edit the endpoint and change the API Key field. If you make the field blank, HyperWhisper deletes the stored key.

Testing an endpoint

The test button is available when you add a new endpoint and when you edit an endpoint. The test sends a minimal request to your server:

macOS
Windows

POST <your endpoint URL>
{ "model": "<your model>", "messages": [{"role": "user", "content": "Say hello in one word."}], "max_tokens": 10, "temperature": 0.0 }

POST <your endpoint URL>
{ "model": "<your model>", "messages": [{"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "Say hello in one word."}] }

Windows sends one system message and one user message. It sends no max_tokens field and no temperature field.

HyperWhisper expects a standard OpenAI-compatible response. The response is a JSON object with a choices[0].message.content field. HyperWhisper saves the last test result (pass or fail) and shows it next to the endpoint name. The result tells you if the endpoint worked at the time of the last test. If you change the endpoint URL, HyperWhisper deletes the test result. The old result does not apply to the new address.

Using a custom endpoint in a Mode

After you save an endpoint, it appears as a provider option in the post-processing settings of a Mode. Select it as you select a built-in AI provider. Read Transcription Modes to learn how to configure post-processing on a Mode.

Managing endpoints

Each endpoint in the list has three actions:

Action	What it does
Edit (pencil)	Changes the name, the URL, the model, or the API key
Duplicate	Makes a copy with a new name. Use this action to try different models on the same server. You do not enter the URL again.
Delete	Deletes the endpoint and its stored API key permanently

When you duplicate an endpoint, HyperWhisper copies the test status with the settings, because the URL and the model are the same. HyperWhisper also copies the API key securely to the new entry. On macOS, if the key copy fails, HyperWhisper deletes the test status. You then know that you must test the duplicate first. On Windows, a failed key copy does not delete the test status. If you think that the key copy failed, test the key of the duplicate manually.

When you delete an endpoint, HyperWhisper removes it from the list and deletes the stored API key. You must change every Mode that used that endpoint to a different post-processing provider.

Common providers

This table gives the base URLs of common OpenAI-compatible services. Use the Custom tab for all of them:

Provider	Base URL
OpenRouter	`https://openrouter.ai/api/v1`
Ollama (local)	`http://localhost:11434`
LM Studio (local)	`http://localhost:1234/v1`
Together AI	`https://api.together.xyz/v1`
Groq (OpenAI-compatible endpoint)	`https://api.groq.com/openai/v1`

For local servers (Ollama, LM Studio), use the Ollama tab or the LMStudio tab. The Custom tab does not find your models automatically.

Endpoint format

A custom endpoint must implement the shape of the OpenAI chat completions API:

POST <your-url>
Authorization: Bearer <api-key>   (omitted if no key)
Content-Type: application/json

{
  "model": "<your-model>",
  "messages": [{ "role": "user", "content": "..." }],
  ...
}

The response must match the standard OpenAI format:

{
  "choices": [
    {
      "message": {
        "content": "..."
      }
    }
  ]
}

Any server that matches this format works: Ollama, LM Studio, vLLM, llama.cpp server, OpenRouter, and more.

Getting Started

Plans & Billing

Transcription & Models

The Interface

Features

Settings

Privacy & Security

Developer & API

Troubleshooting

Windows

Custom OpenAI-Compatible Endpoints

What custom endpoints do

Adding an endpoint

URL format

API key storage

Testing an endpoint

Using a custom endpoint in a Mode

Managing endpoints

Common providers

Endpoint format

​What custom endpoints do

​Adding an endpoint

​URL format

​API key storage

​Testing an endpoint

​Using a custom endpoint in a Mode

​Managing endpoints

​Common providers

​Endpoint format

What custom endpoints do

Adding an endpoint

URL format

API key storage

Testing an endpoint

Using a custom endpoint in a Mode

Managing endpoints

Common providers

Endpoint format