Quickstart

Create a workspace key, verify your connection, and choose your first inference request.

Before You Begin

You need an AudioLab workspace, permission to create API keys, and a terminal with curl. These commands use Bash or Zsh syntax. Model inference may also require available workspace credits and deployment activation for your chosen task.

Open the dashboard, select your workspace, and review its balance in Credits.

1. Create an API Key

Open API Keys, choose Create API key, and give it a name such as docs-quickstart. Set an optional spending limit and expiration, then copy the secret. It is shown only once.

Inference requires the task-specific scope listed in its guide. See Authentication for key permissions and lifecycle management.

Export the key and the Router origin for your environment. The production Router origin is shown below; it does not include /v1:

export AUDIOLAB_API_KEY="YOUR_WORKSPACE_KEY"
export AUDIOLAB_API_URL="https://api.tryaudiolab.ai"

Keep this key on your server. Do not put it in browser code, a public repository, or a client-side environment variable.

2. Verify Your Connection

Request the model catalog:

curl --fail --show-error --silent \
  "$AUDIOLAB_API_URL/v1/models" \
  -H "Authorization: Bearer $AUDIOLAB_API_KEY"

A successful request returns JSON containing a data array of model entries. The list needs no key, but a key you send is checked, so this verifies the Router URL and API key without running model inference. It does not guarantee that every catalog model is enabled for your deployment.

If this request fails, use Errors and Debugging to interpret the status before continuing. A 401 usually indicates a credential problem; 403 indicates an access restriction.

3. Run Your First Inference Request

Choose the task you want to build. Each guide includes the required access, a request example, and instructions for consuming its result:

Task	Guide	Result
Text generation	Text	A message or streamed text
Image generation	Images	Encoded image data
Video generation	Video	A job to poll and download
Speech generation	Text to Speech	Audio bytes
File transcription	Speech to Text	A transcript

Text and embeddings use the same API URL and workspace key. Continue with the Text guide, or choose another task from the task directory.

Keep your first input small. Use an enabled model and the exact request fields in its guide before adding options or automatic routing.

To call the API from code, use the TypeScript and Python SDKs, or point the OpenAI or Anthropic SDK at AudioLab.

4. Inspect the Result

After inference succeeds, open Logs in the same workspace to inspect the request’s status and model. Use Usage to review recorded usage and cost. Retain the request ID when reporting an issue.

A timeout does not establish that the provider did no work. Inspect the request and follow the endpoint’s retry rules before resubmitting. See Errors and Debugging for both failed and uncertain requests.

Once the first request works, continue with Streaming, Model Routing, or an Integration for your application.