Find a voice. Hear the difference.

This page lists the currently selectable shared catalog, including built-in presets and on-demand references. The total and language counts come from the same catalog and availability rules as the console. Browse 24 voices per page or combine language, gender and style filters.

Disabled, withdrawn and rejected references are excluded. Your private workspace clones stay private. Reference playback is evidence of the sample, not proof that every voice has been tested in every synthesis language.

You hear a saved reference recording, not a new synthesis of your input. Click again to pause; switching voices stops the previous sample. Previewing does not consume API balance or homepage generation allowances.

Use the console to generate a short passage of your own. This is the meaningful check for pronunciation, pacing and suitability: the reference sample alone cannot guarantee the result for different text.

Categories describe suggested uses and help you browse; they are not emotion, speed or performance-control parameters. A narration label does not stop a voice being used for product guidance.

Compare two voices using the same text, language and output format. Listen for names, numbers, pauses and the tone your product needs before choosing a default.

Yes, within the 10 configured synthesis languages. The sample describes the voice; the input text and language parameter determine what it says. The website language is only the interface language.

Provide text in the language you want to hear. Chinese text is not translated by selecting English. For very short, ambiguous or mixed-language input, choose the language explicitly or split the text into separate requests.

The voice preset is passed to the console. Sign in if needed, then confirm the selected voice, enter your text and generate. Selection itself does not submit a job or charge your wallet.

Verify your CastReader identity and activate the existing trial without manual approval. Each account receives one trial grant; validity, daily and abuse limits still apply. Then create a server-side API key. Public sample listening needs no activation.

With the CastReader Node SDK, supported built-in aliases such as rowan resolve to a ready default voice. A raw REST request needs the actual voice ID returned by GET /v1/voices for your workspace and processing region.

For a custom clone, save its returned ID after its status becomes ready. Display names are labels, not globally reusable identifiers; do not copy a private ID from a different workspace.

Separate generation from playback. Confirm a successful job, a non-empty audio response and an audio Content-Type before saving it. Saving an error JSON response with an .mp3 extension does not make it audio.

Click the player to satisfy autoplay rules. Download private results through an authorized backend before expiry; never put the permanent API key in an audio URL or frontend code. If the downloaded file also fails, give support the request ID.

Run authenticated generation on your backend with the Node.js or Python client, or use REST from another server language. The browser should call your backend. Permanent API keys must not ship in public JavaScript or mobile binaries.

Migration is not just a base-URL change: voice IDs, authentication, request fields, job polling and audio retrieval differ. Follow the CastReader quickstart and resolve a ready voice in your workspace; another provider’s voice ID cannot be reused.

A new reference may need preparation before it can synthesize. A ready voice can be reused without registration on each request. Preparation time and synthesis queue time are separate; neither is a guaranteed fixed delay.

If preparation fails, inspect the voice error in the console before retrying. For an API permission or region error, confirm the active workspace, key access and resolved region. Do not repeatedly create replacement voices to work around a queue.

You can register your own recording or one with explicit speaker permission covering cloning and your intended use. Public availability of an audio clip is not permission to clone or impersonate its speaker.

Review the voice permissions and service terms for your use, especially redistribution and commercial projects. A sample, library label or successful technical generation is not a substitute for those permissions.