Skip to main content

Voice Providers

Polyphon supports two categories of voice providers: API voices that communicate with a remote model using an API key, and CLI voices that spawn a local command-line tool as a subprocess.

API Providers

Anthropic (Claude)

API key variable: ANTHROPIC_API_KEY or POLYPHON_ANTHROPIC_API_KEY

Available models: Claude Opus 4.6, Claude Sonnet 4.6, Claude Haiku 4.5, and others

To configure:

  1. Get an API key from console.anthropic.com.
  2. In Polyphon, go to Settings → Anthropic and paste your API key.
  3. Click Test to verify the key is working.

Anthropic provider card expanded showing API key status and voice type selector

Alternatively, set the key in your shell environment (~/.zshrc or ~/.bash_profile) as ANTHROPIC_API_KEY. Polyphon reads your login shell environment at startup, so keys set there are automatically picked up.

OpenAI (GPT)

API key variable: OPENAI_API_KEY or POLYPHON_OPENAI_API_KEY

Available models: GPT-4o, GPT-4o Mini, GPT-4 Turbo, and others

To configure:

  1. Get an API key from platform.openai.com.
  2. In Polyphon, go to Settings → OpenAI and paste your API key.
  3. Click Test to verify.

Google (Gemini)

API key variable: GOOGLE_API_KEY or POLYPHON_GOOGLE_API_KEY

Available models: Gemini 2.0 Flash, Gemini 1.5 Pro, Gemini 1.5 Flash, and others

To configure:

  1. Get an API key from aistudio.google.com.
  2. In Polyphon, go to Settings → Google and paste your API key.
  3. Click Test to verify.

CLI Providers

CLI voices run a local command-line tool as a subprocess. No additional API key configuration is needed in Polyphon — the CLI tool manages its own authentication.

Claude CLI (claude)

The Claude CLI is Anthropic’s official command-line interface.

Requirement: claude must be in your system PATH and authenticated.

To verify it is available:

claude --version

In Polyphon, go to Settings → Claude CLI and click Test to confirm Polyphon can find and invoke it.

Claude CLI provider card with Available status indicator

OpenAI Codex CLI (codex)

Requirement: codex must be in your system PATH and authenticated.

codex --version

GitHub Copilot CLI (copilot)

Requirement: copilot must be in your system PATH and authenticated via the GitHub CLI.

gh copilot --version

API Key Resolution

Polyphon resolves API keys in this order:

  1. A provider-specific Polyphon variable (e.g. POLYPHON_ANTHROPIC_API_KEY)
  2. The provider’s canonical variable (e.g. ANTHROPIC_API_KEY)
  3. A key saved through the Settings UI

API keys are resolved in the main process only and never sent to the renderer or logged. The Settings page shows only a masked representation of the key.

Checking Provider Status

The Settings page shows the status of every configured provider at a glance:

  • Available — the key or CLI tool was found and tested successfully
  • Key found (untested) — a key is present but has not been tested
  • Not configured — no key or CLI tool found

Provider settings page showing multiple provider cards in different states

Custom OpenAI-Compatible Providers

In addition to the built-in providers above, Polyphon supports custom OpenAI-compatible endpoints. Use this to run local models with Ollama or LM Studio, connect to a private inference proxy, or use any service that exposes an OpenAI-compatible API — without modifying Polyphon.

Custom providers are configured in Settings → Custom Providers. Once added, they appear in the Composition Builder voice selector alongside the built-in providers listed above.

See Custom Providers for step-by-step setup instructions, including an Ollama example.

Settings Providers tab showing all provider cards with enable toggles