Terms of Service

Last updated

These Terms of Service (the "Terms") are an agreement between you and ("we", "us") covering the website, console and application programming interface (together, "Ciabatta" or the "Service"). By creating an account, creating an API key or using the Service, you agree to these Terms, our Privacy Policy and our Acceptable Use Policy. If you use the Service for an organization, you accept these Terms on its behalf and confirm you have authority to do so.

1. Accounts and API keys

• You must be at least 18 years old and able to form a binding contract.
• You are responsible for everything that happens under your account and your API keys. Keep keys on your server, never embed them in apps or web pages, and revoke any key you believe is exposed. Give browsers and mobile apps only the short-lived session tokens the API issues.
• Provide accurate information and keep it current. One free signup credit per person or organization; creating accounts to collect additional free credit is prohibited.
• Related products. We also operate other speech products that use the same sign-in system. Your sign-in, API keys and credit balance work across them, and each product's own terms apply when you use that product. These Terms alone govern your use of Ciabatta.

2. The Service

Ciabatta lets you build voice agents: it transcribes audio you or your end users stream to it, generates replies with large language models, optionally calls tools you define, and speaks the replies with synthetic voices, including voices you clone. It also offers direct access to those building blocks (chat completions, speech-to-text and text-to-speech). We may change, add or remove features. We will give at least 30 days' notice by email before removing an API endpoint, parameter or curated model you have used in the previous 90 days, except where a change is required for security or legal reasons or the model provider retires the model sooner, in which case we will move your agents to the closest equivalent model and tell you.

3. Experimental models

You may opt in to experimental models that are not on our curated list. Experimental models come with no latency, quality or tool-calling commitments, may produce lower-quality or unexpected output, and may be removed with days' notice. They are billed as described on the pricing page.

4. Credits, pricing and payment

• Prepaid credit. The Service is pay-as-you-go from a prepaid dollar balance on your account. Usage is charged at the prices shown on our pricing page and returned by GET /v1/models at the time the usage happens.
• Free signup credit. New accounts may receive a one-time promotional credit. It has no cash value and cannot be transferred or refunded.
• Per-turn pricing. Voice agents are charged per turn (one user utterance and the agent's reply, including any tool calls), at the price for the model tier when the turn starts. Each turn includes allowances for prompt and output tokens, spoken characters and connected time. Usage beyond an allowance, idle connected time, text-only turns, memory summaries, direct speech requests and experimental models are charged at the listed rates. Every usage report shows allowance consumption so you can see any extra charge.
• Reservations. We reserve the maximum possible cost of a turn from your balance when it starts and charge only the actual cost when it ends. If your balance runs out, the session ends.
• Interrupted and failed turns. A turn that is interrupted by the user, cancelled by you or disconnected after it started is charged for the work done. Turns and requests that fail because of an error on our side are not charged, and any reserved amount is returned to your balance automatically.
• Spend caps. Spend caps and alerts you set are applied on a best-effort basis and can be exceeded by up to the cost of the turns already in progress.
• Auto top-up. If you turn on auto top-up, you authorize us to charge your saved payment method the amount you chose whenever your balance falls below the threshold you set, up to the monthly limit you set. You can turn it off at any time in the console.
• Refunds. Purchased credit does not expire while your account is open. Unused purchased credit is refundable within 30 days of purchase on request to ; after that, credit is non-refundable except where required by law. Credit is forfeited if your account is deleted or terminated for breach of these Terms.
• Price changes. We will give at least 30 days' notice before raising a price. Credit is a dollar balance, so price changes do not reduce credit you have already bought.
• Chargebacks. If you dispute a payment, we may suspend the account until the dispute is resolved.
• Taxes. Prices exclude taxes unless stated. You are responsible for applicable taxes.

5. Your content and your end users

• "Your Content" means agent configurations, prompts, tool definitions, reference recordings, audio and text you or your end users send, and the transcripts, replies and audio the Service produces for you. You keep all rights to Your Content and may use the Service's output for any lawful purpose, including commercially.
• You grant us a limited license to process Your Content only as needed to provide, secure and support the Service, as described in the Privacy Policy. We do not use Your Content to train models.
• You are responsible for having all rights needed to submit Your Content and for how you use the output.
• You are responsible for your end users. For personal data of the people your agents talk to ("End Users"), you are the controller and we process it on your behalf. You are responsible for giving End Users any required notices, obtaining any required consent (including all-party consent to recording or transcription where the law requires it), honoring their requests, and choosing a retention setting that fits your obligations.
• AI disclosure. You must tell End Users they are talking to an automated system wherever the law requires it or where they could reasonably believe they are talking to a person.
• Identifiers. Use opaque identifiers for customer_id. Do not use email addresses, phone numbers or other direct identifiers.
• AI output can be wrong, incomplete or inappropriate. You are responsible for reviewing how your agents behave and for not relying on them where a mistake could cause harm (see the Acceptable Use Policy).

6. Voice cloning

Voice cloning is available only after you accept our Voice Cloning Terms in the console. Your acceptance, the version you accepted, the date and time, the name you typed and a hashed record of your network address are stored with your account. In summary, and without limiting those terms:

• You may only clone your own voice, or the voice of a person who has given you clear, informed, written permission that you can produce on request.
• You must not clone a public figure, celebrity or performer without documented authorization, or a minor without verifiable consent from a parent or legal guardian.
• You must not use a cloned voice, in an agent or otherwise, to impersonate, deceive, defraud or harass anyone, to fabricate statements by a real person, to interfere with elections, or to defeat voice-based security.
• You are solely responsible for the voices you clone and the audio your agents speak with them, and you will indemnify us against claims arising from that use.

We may remove cloned voices, disable agents that use them and suspend accounts on a credible report of misuse. Report misuse on our abuse page.

7. Tools and webhooks

When your agent calls a tool, we send the request to the URL or client you configured, on your behalf and with the arguments the model produced. You are responsible for your tool endpoints, for validating every request (including its signature), and for any action they take. Model-generated arguments can be wrong; do not let a tool take an irreversible or high-impact action without your own checks.

8. Acceptable use

You must follow the Acceptable Use Policy, which includes the use restrictions of the models we build on. You must not reverse-engineer the Service, attempt to extract model weights or other customers' voices or data, use multiple accounts to get around limits or enforcement, resell raw API access as a competing voice agent or speech service without our written permission, or probe or disrupt our systems.

9. Suspension and termination

You may delete your account at any time from the Account page of the console. We may suspend or terminate access immediately if you breach these Terms, if required by law, or to protect the Service, End Users or others. Where reasonable, we will tell you why and give you a chance to fix the problem. Sections 5, 6, 7, 11, 12, 13 and 14 survive termination.

10. Availability

We work hard to keep the Service fast and available, but it is provided without a service-level agreement unless we sign a separate agreement with you. Latency figures we publish are measurements, not guarantees. Scheduled maintenance will be announced in advance where practical. Ciabatta is not designed for emergency services, and you must not rely on it to place or answer emergency calls.

11. Disclaimers

THE SERVICE AND ALL OUTPUT ARE PROVIDED "AS IS" AND "AS AVAILABLE". TO THE FULLEST EXTENT PERMITTED BY LAW, WE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, NON-INFRINGEMENT AND ACCURACY. Transcription, language model replies and synthetic speech can all contain errors.

12. Limitation of liability

TO THE FULLEST EXTENT PERMITTED BY LAW, WE WILL NOT BE LIABLE FOR ANY INDIRECT, INCIDENTAL, SPECIAL, CONSEQUENTIAL OR PUNITIVE DAMAGES, OR LOST PROFITS, REVENUE OR DATA. OUR TOTAL LIABILITY FOR ANY CLAIM RELATING TO THE SERVICE IS LIMITED TO THE AMOUNT YOU PAID US IN THE 12 MONTHS BEFORE THE EVENT GIVING RISE TO THE CLAIM.

13. Indemnity

You will defend and indemnify us, our affiliates and our personnel against claims, losses and costs (including reasonable legal fees) arising from Your Content, your agents' conversations with End Users, your cloned voices, your tool endpoints, your failure to give notices or obtain consent, or your breach of these Terms.

14. Governing law and disputes

These Terms are governed by the laws of , without regard to conflict-of-law rules. Before filing a claim, each party agrees to try to resolve the dispute informally for 30 days by contacting the other in writing.

15. Changes to these Terms

We may update these Terms. For material changes we will email you at least 30 days before they take effect. Continuing to use the Service after that date means you accept the updated Terms.

16. Contact

Questions about these Terms: .