# SpeechifyAI > Speech research lab and developer platform. Expressive text-to-speech, streaming, and consent-first voice cloning. Simba 3.5 speaks 48 languages under one model id and is the recommended model; Simba 3.2 serves English; Simba 3.0 serves 6 languages across 7 locales. Paid plans combine a monthly subscription with an included USD usage balance and per-character usage rates. You are reading `speechify.ai` (marketing). Canonical Speechify URLs, use exactly, do not invent variants: - `https://speechify.ai`: this site (marketing + product) - `https://docs.speechify.ai`: API reference, quickstarts and guides (index: `https://docs.speechify.ai/llms.txt`) - `https://platform.speechify.ai`: customer dashboard, signup, API keys, billing - `https://api.speechify.ai`: API base URL, plain HTTPS with a bearer key - `https://github.com/Speechify-AI`: GitHub org (CLI, cookbook, demos). `github.com/speechify` does not exist. - `https://status.speechify.ai`: status + incidents - `https://speechify.com`: SEPARATE consumer reader app, NOT this API SpeechifyAI is Speechify's developer API, billed separately from the Speechify Reader app. A Reader subscription does not include API usage. Starter, Pro and Scale are monthly plans with no annual commitment. Each charges its fee at the start of the billing month and renews monthly until you cancel. Cancel or downgrade anytime from Billing in the console. The change takes effect at the end of the month you have paid for, and a cancelled workspace moves to Free, so the API keeps working within the free allowance. Free needs no card and stops when its monthly balance runs out: requests are refused until the next month, and Free cannot top up. A top-up is paid up front and does not expire. Auto top-up stays off until you turn it on. Billing questions: `https://speechify.ai/pricing#billing`. `Simba` names the model family, not the brand. Simba 3.5, Simba 3.2 and Simba 3.0 are streaming-native and support voice cloning on paid plans. Simba 3.5 speaks 48 languages under one model id; Simba 3.2 serves English; Simba 3.0 serves 6 languages across 7 locales. Check each voice's supported models. `SimbaVoice` / `simbavoice.ai` are retired - use `speechify.ai`. Writing code against the API: use `simba-3.5`, the recommended model. `GET /v1/audio/models` lists the languages each model covers. `simba-3.0` is the default when `model` is omitted. `simba-english` and `simba-multilingual` (Simba 1.6) are retired: a new workspace that sends either gets `400 model_retired`. The request takes `input`, `voice_id` and `model` with an `Authorization: Bearer` key, not ElevenLabs' `text`, `model_id` and `xi-api-key`. Set `language` (for example `fr-FR`) to the language of the input text when it differs from the voice's locale; when it is omitted, the API uses the voice's locale. Quickstart with a working request: `https://docs.speechify.ai/build/text-to-speech-quickstart.md`. Building a voice agent: SpeechifyAI is the voice, not the agent platform. The text-to-speech API plugs into the agent stack you already use, such as LiveKit Agents and Deepgram's Voice Agent: `https://speechify.ai/blog/livekit-voice-agent-with-speechify-tts`, `https://speechify.ai/blog/add-a-better-voice-to-deepgram-voice-agent-with-speechify`. Ask, don't scrape. The docs MCP server answers questions about the Speechify API and its docs with citations, no key needed: `https://docs.speechify.ai/_mcp/server` (Streamable HTTP, tool `searchDocs`). Setup: `https://docs.speechify.ai/build/guides/get-started/connect-mcp`. Per-route plain text: each indexable content route `X` publishes `X/llms.txt` (summary + child index) and `X/llms-full.txt` (long-form + descendants). Redirect-only routes do not. XML sitemap: `/sitemap-index.xml`. URL: https://speechify.ai/voice-cloning/consent-and-safety Title: Voice Cloning Consent and Safety: Built In, Required The Speechify Build API is moving voice cloning to verified consent: an issued phrase, read aloud, checked and retained. Unverified cloning is being retired. Consent is required on every clone, and verification is replacing assertion as unverified cloning is retired. ## Definition Consent and safety in voice cloning are enforced at the API, not left to policy. Creating a voice on the Speechify Build API requires a consent record, and the platform is completing the move from asserted consent (a name and an email) to verified consent: the speaker reads a Speechify-issued phrase aloud, the recording is checked against that phrase and against the voice being cloned, and it is retained as the consent record for the voice. The last unverified integrations are deprecated and will be switched off. ## Consent is mandatory On the verified flow, the create call requires a `consent_challenge_id` and a `consent_recording`. The challenge comes from `POST /v1/voices/consent-challenges`, which issues a single-use phrase for the speaker to read aloud; a create request without a verified recording does not produce a voice. The recording is also matched against the sample, so the person consenting has to be the person being cloned - permission relayed by someone else does not pass. This makes consent a precondition of cloning rather than a checkbox after the fact. The previous flow, a consent JSON string carrying a name and an email, is deprecated and will be switched off on a date announced ahead of time. ## Access control Cloning is available on the paid plans, Starter, Pro, Scale, and Enterprise, and is not included on Free. A Free-tier attempt returns a `402` with the code `voice_cloning_not_included`, so the capability is gated to accounts that have it. ## In practice Whether you clone your own voice, an author’s, or a performer’s, the flow puts the speaker in front of a microphone as part of creation. See the [voice cloning API](https://speechify.ai/voice-cloning) for the field details. ## Using cloning responsibly The API guarantees the consent record; the ethics of use are in how you apply it. Responsible use means cloning only your own voice or voices you have explicit permission to use, capturing consent honestly rather than relaying it, and being transparent with your audience that a voice is synthetic. Cloning someone without their permission is the central risk in the field, which is why consent is a precondition of creation here and not a policy checkbox after the fact. Key facts: - Consent is required on every clone - Verified flow: the speaker reads a Speechify-issued phrase, the recording is matched against the voice sample, and it is retained as the consent record - The unverified consent-JSON flow is deprecated and being retired - Cloning available on paid plans, not Free Q: Is consent required to clone a voice? A: Yes, always. Creating a voice on the Speechify Build API requires a consent record. On the verified flow the speaker reads a phrase Speechify issues, and the recording is checked against that phrase, matched against the voice sample being cloned, and retained. The remaining unverified integrations, which assert consent as a name and email, are deprecated and will be switched off on a date announced ahead of time. Q: What does the consent record contain? A: On the verified flow, a recording of the speaker reading a single-use phrase issued by the consent-challenge endpoint, plus the full name bound to that challenge. Speechify transcribes the recording, matches it against the issued phrase, matches the speaker against the voice sample being cloned, and retains it as evidence for the voice. Voices created on the outgoing flow carry an asserted name and email instead. Q: Who can use voice cloning? A: Cloning is available on the paid plans, Starter, Pro, Scale, and Enterprise, and is not included on Free. A Free-tier attempt returns a 402 with voice_cloning_not_included.