API Terms
These terms cover programmatic access to Timbrica tools. They add to the Terms of Service and take precedence over them for anything API-specific. Using the API means accepting them.
Two things carry most of the weight here, so they are stated up front: you are responsible for every call made with your key, and you are responsible for telling your end users that a result came from AI. Everything below elaborates.
1. Keys and access
A key is created by a human in the account settings of a registered account. Automated account or key creation is not supported and not permitted.
The key is shown once. We store only an irreversible hash and cannot recover a lost key — mint a new one and revoke the old.
Every run made with your key counts as yours, whether a person or an agent initiated it, including the tokens it spends. Keep the key out of client-side code, public repositories and shared configuration. If you suspect it leaked, revoke it in the account — that takes effect immediately.
Do not hand your key to a third party. If you build on top of the API for others, you contract with them yourself; they never act under your key as if it were theirs.
Test keys return stub results and spend nothing. Use them while building. They are not a way to obtain usable results without paying.
We set per-key limits — request rate, concurrent runs, daily token ceiling, allowed tools, allowed IPs. Current values are visible in the account and through the API.
2. Paying
API usage is paid in tokens from the same wallet as the site. There is no separate access fee and no subscription.
A run's price can be quoted with
POST /v1/tools/{slug}/estimate before anything is
spent. The price is frozen when the run is accepted and never changes afterwards.
Tokens are charged only for a delivered result. A technical
failure, a cancellation or an expired run releases the hold in full. A retry with the same
Idempotency-Key returns the stored result and does not charge
twice.
Refunds for token purchases follow the Refund Policy. Tokens already spent on delivered results are not refundable except where that policy or applicable law says otherwise.
3. Your end users
We have no contract with your end users and owe them nothing. You act in your own name and answer to them for your product.
You must pass on to them what affects their rights: that AI is involved (§5), the prohibitions in §4, and how their data is handled (§6).
You may build the results into your product, including commercially. You may not present our API, tools or infrastructure as your own, and you may not make guarantees on our behalf that we do not make ourselves (§8).
4. What you must not do
Through the API you must not:
- Synthesise, clone or imitate the voice of a specific real person without their explicit consent, or produce audio implying that a specific person said something they did not.
- Create or alter images or video of a specific real person in a way that creates a false impression of authenticity, including face replacement without the depicted person's consent.
- Create intimate material depicting a real person without their consent, or any sexual material involving minors or their likeness. This one has no exceptions and no grace period.
- Mislead about origin: strip or hide AI-provenance markers (§5), pass generated material off as human-made, or imitate communications from governments, news organisations or companies.
- Work around limits: multiple accounts or keys to exceed ceilings, spreading load for the same purpose, circumventing regional tool restrictions, or scraping the site to exceed what the API allows.
- Send bulk unsolicited messages, distribute malware, attack systems, or do anything unlawful in your jurisdiction or ours.
- Submit material you have no right to submit.
Your end users' compliance is your responsibility. We do not pre-screen request content and are not obliged to.
5. AI disclosure
Telling end users that material was created or altered by AI is your obligation, because you are the one facing them. Where Article 50 of Regulation (EU) 2024/1689 (the AI Act) applies to you or to them, discharging it is on you.
In formats that support it, we embed a machine-readable AI-provenance marker in the result's metadata. Removing or altering it to pass the material off as human-made is prohibited (§4).
We do not verify your compliance with this section and are not liable for it to third parties.
6. Data
For end-user data you send us, you are the controller and we are the processor; this section is the processing instruction. You must have a lawful basis for sending it, and you warrant that you do.
We process what you send solely to execute the run. We do not use it to train models and do not disclose it except where the law requires. Our security and confidentiality commitments are in the Privacy Policy.
Where it is processed: a key belongs to the host it was
created on. Calls to tembrica.com are processed in the
European Union; calls to timbrica.com are processed in Russia.
Presenting a key to the wrong host is rejected with
421 and the correct base URL — deliberately, so data never
crosses shards by accident.
Input files are deleted within 4 hours of upload; results within 24 hours unless a tool states otherwise. Collect your results inside that window.
We keep operational records of runs — time, tool, size, price, status — for accounting and dispute resolution. Request content is not kept in them.
7. Suspension and revocation
We may suspend a key immediately on signs of compromise, abnormal spend, sustained failures or a breach of these terms. Suspension is reversible and we notify the key's owner. An automatic suspension is cleared by the owner in the account once the cause is fixed.
For a serious or repeated breach of §4 we may revoke the key and decline further API access. Your remaining token balance stays yours and remains usable on the site.
8. Availability
The API is provided as is. We do not guarantee uptime, latency or fitness for your purpose, and we do not offer an SLA.
Build for failure: handle errors, retry with an idempotency key, and do not let our downtime lose your users' data.
9. Contract versions and changes
Every tool contract is versioned and its history is published in the changelog. Compatible changes — new tools, new fields, new optional parameters — ship without notice.
Breaking changes get at least 180 days' notice before they take effect. The exception is a change forced by law, by a rightsholder, or by a security vulnerability, which may take effect immediately.
We may change these terms. A new revision is published on this page with its date; continuing to use the API after it takes effect means accepting it.
10. Liability
You indemnify us against losses arising from your breach of these terms, including third-party claims relating to your or your end users' use of the API.
Our liability to you is capped at the value of tokens you spent through the API in the three months preceding the event giving rise to the claim.
Nothing here limits consumer rights that cannot be limited by agreement under the law of your country of residence.
11. Contact
API questions, key suspensions and abuse reports: faniazaripova2023@yandex.ru.
Revision of 2 August 2026.
