Overview

AllModels offers a single API endpoint that routes text‑to‑speech, speech‑to‑text, and real‑time voice models from multiple providers. By normalizing each provider's schema, you can call a unified endpoint instead of dealing with individual APIs, which simplifies both development and maintenance.

The platform currently exposes 91 model routes across 18 different voice providers. Each request is routed with a latency of less than 30 ms, and the average latency for a typical call is 87 ms (50th percentile) and 214 ms (95th percentile). You can target a route by selecting criteria such as latency, cost, or specific capabilities. For example, the route “cartesia/sonic‑3.5” serves a TTS application and can be chosen through the API by providing a JSON payload that identifies the desired model. The API accepts simple JSON request bodies and returns JSON responses, making integration into existing systems straightforward.

Voice portability is a key feature. If you want to keep a user's preferred voice when switching providers, you send the source provider and voice name, select a destination provider, and the API returns a compatible voice slug. The example shows sending a POST to /v1/voices/port with `{ "source": { "provider": "elevenlabs", "voice": "edward" }, "destination": { "provider": "cartesia" } }`. The response includes `{ "voice": "elevenlabs/edward", "provider": "cartesia", "status": "ready" }`. This slug can be used in compatible routes such as cartesia/sonic, allowing you to migrate voices while preserving the user experience.

Account management is handled through the MCP (Managed Client Platform) client. The MCP client can query balance information, create top‑up links, and manage guardrails or API keys. Authentication is performed via OAuth on the AllModels dashboard, ensuring no API key is stored in the client. To add a connected client from a command line, you can use `claude mcp add --transport http allmodels https://mcp.allmodels.io/mcp`. Functions such as `get_balance` return credit usage and USD balance, for example: `{ "credits": 42500000, "balanceUsd": 42.5 }`.

AllModels is ideal for developers who need to aggregate multiple voice models into a single interface, especially when latency or cost considerations are important. It may be less suitable for teams that require detailed pricing breakdowns or the ability to access provider‑specific advanced features that are not exposed through the unified schema. The directory listing does not provide explicit pricing information or a complete list of model capabilities beyond the latency statistics.