Terms
Version v2026-09 · Effective 2026-11-11
These Terms govern your use of avidol (the website, the Director Workspace, the API, the remote MCP server, the official Skill and CLI, and the Telegram partner bot — the "Service"). The Service is provided by Avidol.ai ("we"). By using the Service you agree to these Terms; if you do not agree, do not use it.
1. Eligibility and age
You must be at least 18 to use the Service.
The Director Workspace does not currently ask you to tick a separate age declaration, and it does not split content into ratings: self-registered accounts can use characters of every rating. When you upload material you must confirm that everyone in it is at least 18.
2. Accounts, sign-in and API keys
You can sign in with an email sign-in link (valid for 15 minutes and usable once) or with a Google account; we do not offer Telegram sign-in. A sign-in session lasts 30 days.
An API key is shown once when it is created, and we keep only its hash and a few leading and trailing characters. Keep it safe: any request made with your key is treated as your own action. After you rotate a key, the old key keeps working for a 24-hour grace period and then stops. You can revoke a key at any time.
You can set hourly, daily or lifetime credit limits for each key; a read-only (report) key can read records but cannot place orders.
When you first sign in and create your account, you are asked to tick a box accepting these Terms (it is never pre-ticked); later sign-ins, orders and changes to these Terms do not ask you to tick it again and do not restrict your account. By accepting, you warrant that you will comply with the laws of the place where you are located and that you are responsible for confirming that using the Service and the generated content is lawful there. We record the version you accepted, the time, and the country or region code of your connection at that moment; we do not restrict your use of the Service based on where you connect from. An API-only account can also record its acceptance through the account-update API (PATCH /v1/me with the Terms version).
If you link the Telegram partner bot with a one-time invite code (valid for 7 days), commands sent through the bot (for example placing an order) are treated as actions of your account and follow the same quote, hold and refund rules as the Workspace.
You must keep your sign-in methods and API keys secure, and you are responsible for everything done under your account and keys, except where we are at fault. If a key leaks or someone else uses your account, revoke the key immediately and write to support@avidol.ai.
3. Credits, top-ups and receipts
The Service is priced in credits (✦). Each US$1 you top up gives 100 ✦, so 1 ✦ corresponds to US$0.01 of top-up principal. A single top-up is between US$5 and US$5,000 (US$20 to US$5,000 for USDT), with at most two decimal places. Prices are set at face value, and credits are always added at face value. A USDT top-up is paid at face value (for example, ✦2000 costs US$20.00); for Workspace and API accounts paying on the sidediff checkout page (for example with Alipay), you pay the face value × 1.2 (for example, ✦2000 costs US$24.00), so paying with USDT saves about 17% compared with Alipay; top-ups on the viewer site are currently paid at face value. The part of a payment above the face value is not refundable except where, for reasons attributable to us, we cannot provide a service you paid for, where you were charged twice or in error, or where the law requires it to be returned (in those cases we return the amount corresponding to what was affected, with a calculation).
Top-ups carry no bonus credits. Bonus credits already credited before this change are not taken back and remain usable, but they cannot be withdrawn or exchanged back into cash.
The payment methods available are those shown on the top-up page at the time. They currently are: Alipay (paid on the checkout page of the third-party payment provider sidediff, charged in Chinese yuan) and USDT (via the third-party payment provider NOWPayments, priced in US dollars; please transfer on the TRC20 network, and the on-chain network fee for sending is borne by the payer). When you pay on a provider's page, your payment details are entered on that page; we do not receive or store them.
For Alipay the amount is converted at the exchange rate in effect when the order is created and rounded up to a whole yuan; that rate is recorded on the top-up. The rate is updated daily as the highest US-dollar selling rate published by three Chinese banks plus 2.5%; if it has not been updated for 72 hours, Alipay is suspended.
Credits are added according to the payment status and amount we look up directly with the provider: only a successful payment is credited; a failed or cancelled payment is not; an order left unpaid for 24 hours expires automatically; and a payment notice delivered more than once is not credited twice. If you pay on the sidediff checkout page and less than the amount due is paid, the original order is voided and a new one is credited for the amount actually received: the amount received (converted into US dollars) is divided by the multiplier stored on that order when it was created (1.2 for Workspace and API accounts) and rounded down to the cent, so if US$12.00 was due and US$11.00 arrived, ✦916 is credited; if more is paid, only the original order's credits are added; if the currency or amount cannot be matched, nothing is credited automatically and we handle it manually. A USDT top-up is credited only once the provider confirms the payment as complete and both the currency and the amount match; a partial payment or a mismatched amount is not credited automatically, so please contact support@avidol.ai.
You may receive sign-up bonus credits on your first sign-in; the amount is set by us in the back office and may be zero. We may also grant trial or promotional credits. Granted credits cannot be withdrawn and cannot pay for Standard Cloud.
For each credited top-up you can issue one receipt from your wallet, numbered "R-year-six-digit serial". There is only one receipt per top-up and its content cannot be changed once issued. A receipt is not an invoice. The issuer details on a receipt are taken from the platform settings at the time it is issued; any detail not set is left blank.
The system does not currently deduct credits because time has passed since a top-up; this does not mean credits can be withdrawn or exchanged for cash. If we introduce a validity period in future, we will give advance notice as applicable law requires and explain how existing balances are treated.
4. Quotes, holds and settlement
A quote does not take credits. Credits for the quoted amount are held only after you confirm, and the quoted amount is charged when the deliverable is delivered. The billing terms for each job are shown at quote time and are never changed retroactively. A re-render is a new paid quote.
How a failed or cancelled job is billed follows the billing terms written into that job at quote time, which the quote screen shows. Jobs created in the Director Workspace always use "incurred-cost billing" (segment re-renders of a multi-segment job keep the original job's terms): failed or cancelled jobs (including deliverables withheld because of a minor determination) are charged the verified upstream cost (converted to credits rounded up and capped at the quote), with the rest returned; if the verified cost is zero, everything is returned. Orders placed through the API, the remote MCP server, the official Skill and CLI or the Telegram partner bot without specifying billing terms, and single-shot re-renders requested through the API (wherever the original job was created), use "no output, no charge": if generation fails or is cancelled, the held credits are returned in full.
While the cost is unverified, held credits stay on hold pending reconciliation; this does not mean they were charged or refunded. If the cost is still unverified seven days after the job's last update, everything is returned. Check your transaction history for the actual charges, refunds and pending items; a failed or cancelled status alone does not mean a full refund.
A queued job can be cancelled; once generation starts it cannot be; a Standard Cloud job cannot be cancelled once it has been sent to the upstream provider (even if the screen still shows it as queued).
Each account has a limit on jobs running at the same time (2 by default); jobs above the limit are deferred. We can set a monthly consumption cap for an account; once it is reached, new holds are refused.
If the Director Engine has no machine available, the quote fails; you can instead send a "queue request", which takes and holds no credits, and we email you when a machine is available so you can come back and quote yourself. Each account can have at most 5 queue requests at a time.
Text-model calls are billed differently: before a call, credits are held up to the approved usage limit; when it completes, it is settled at the actual usage reported by the upstream provider and the price approved at the time (never more than the hold) and the rest is returned; a call that expires before it is sent is returned in full once we confirm it. A call whose sending or outcome cannot be confirmed is pending reconciliation and is not treated as having incurred cost: we check with the provider manually, settle it at the actual usage if usage is confirmed, and return it in full if no usage is confirmed; the rule above that returns everything after seven days does not apply to such calls. An automatic release for checks that produce no answer has not been built yet; until it is, the text-model interface is available only to accounts we approve request by request.
5. Generation engines and upstream providers
The "Director Engine" runs on GPU machines we rent from Vast.ai, and you can order only when a machine is available; when none is available the quote fails and no credits are taken.
"Standard Cloud" resells third-party generation services: ReachAPI (the MiniMax H3 model) and BytePlus ModelArk (the Seedance 2.0, 2.0 Fast and 2.5 models), priced as shown in the public price list and on the quote; once you confirm a quote, its amount is not affected by later price changes. Your prompt and the reference images of the chosen character look (provided as signed links valid for 15 minutes) are sent to the upstream provider: ReachAPI receives the first two, and BytePlus (Seedance) receives only the first: the Seedance 2.0 models use it as the video's first frame, while Seedance 2.5 uses it as a reference image, so it is not guaranteed to become the first frame.
Upstream providers have their own terms of service and usage policies, which also apply to work sent through Standard Cloud and to text-model calls processed by ReachAPI. When a provider changes its service, models, prices or policies, the models, parameters and prices available on Standard Cloud may change accordingly; quotes you have already confirmed are not affected. The ADULT label does not widen what an upstream provider allows; content prohibited by an upstream provider's policies must not be submitted even if ADULT is selected. The upstream terms and policies that currently apply include the ReachAPI Terms of Service (https://reachapi.ai/terms-of-service) and the BytePlus GenAI Acceptable Use Policy (https://docs.byteplus.com/en/docs/legal/acceptable_use_policy_byteplus_genai); these addresses were checked on 27 October 2026. Upstream terms apply only so far as they relate to your use of the model you choose, and they do not reduce the rights you have against us under these Terms or applicable law; if a new upstream restriction makes a job that has not yet run impossible to run, the credits held for it are handled under section 4 and this section.
Whether Standard Cloud accepts a job is decided by the upstream provider under its own rules. If the provider rejects a job when it is submitted, the job ends immediately and the screen shows "provider rejected" (error code PROVIDER_REJECTED); the provider's original error code is kept in our audit records and is not shown on screen. A rejection may be due to the content or the parameters, or to the status of our account with the provider (for example, exhausted quota); when the cause is on our side, the screen shows "service temporarily unavailable" (PROVIDER_UNAVAILABLE), which does not mean your content broke any rule, and the held credits are returned in full. Held credits: if ReachAPI rejects the job, they are returned in full at once; if BytePlus (Seedance) rejects it, a job under "no output, no charge" gets them back in full at once, while a job under "incurred-cost billing" stays pending reconciliation because the upstream cost is not yet verified, and everything is returned if the cost is still unverified seven days after the job's last update.
To use Standard Cloud, your credited cash (live top-up amounts actually received, net of the fees the payment provider actually charges, plus earnings from your works and any allowance we separately approve) must cover the upstream cost; granted, bonus and sandbox credits do not count as cash.
Video length depends on the model: Seedance 2.5 offers 5 to 30 seconds in 5-second steps; other models offer 5, 10 or 15 seconds. The two engines accept different parameters; see "Which parameters each engine accepts" in the API documentation. The output quality check (quality_check report/enforce) and assembling segments into one video (assemble) are not open yet; multi-segment jobs are delivered segment by segment.
The API also has a text-model call interface, which can be used only after we approve and enable each request.
6. Your content and generated content
You must have the right to use what you upload and the prompts you write. When uploading you must confirm that everyone in the material is at least 18, that you hold the likeness rights, and that you accept the current upload rules.
Content involving minors is prohibited, whichever engine you use. At quote time, a prompt that contains words describing minors or an age under 18 is refused (no job is created and no credits are taken).
The Director Engine worker samples frames and runs facial age estimation when it produces a deliverable; a deliverable judged to show a minor or a possible minor is not delivered, and credits are handled under section 4. If that check could not run, the deliverable is delivered as usual. Standard Cloud deliverables are produced by the upstream provider, and we do not run this check on them before delivery.
Uploads are not reviewed by us before use, except that reference images for a private character look must first pass a source review, in which authorised operations staff view the image and the source evidence you provide (see section 3 of the Creator Agreement). Publication review happens only when you ask to publish a work on the avidol platform. Other prohibitions are in the Content Policy; content involving minors is handled under Reporting and Takedown.
We do not currently offer a data processing agreement (DPA). A business customer that needs us to process personal data of its own customers or other third parties on its behalf must contact us first and put an applicable data processing arrangement in place; until then, please do not submit such data for that purpose.
7. Deliverables, rights and retention
Regular deliverables are kept until the later of 30 days after the job's creation and 90 days after your most recent credited top-up, then deleted automatically; the extension from top-ups counts toward your account's storage quota, and when the quota is exceeded the oldest extended deliverables fall back to their original retention date (normally 30 days after creation); intermediate candidate files produced during generation are kept for 7 days. Download what you need within that time.
Your uploads are kept for 30 days from upload and then removed automatically, with two exceptions: (1) uploads used in your scenes or private looks (your library) are kept as long as your account is active and used within the last 360 days, and fall back to the 30-day rule once the account is suspended or idle for more than 360 days, or once that private look (or its character) is retired; (2) material used by a published work that passed review, and publishing on the platform is not open yet. You can also delete uploads that no job is currently using.
As between you and us, we claim no rights in the materials you upload, your prompts or your deliverables. You grant us permission to store, process and transmit that content to the extent needed to provide, maintain and protect the Service (including sending it to the upstream providers listed in section 5); this permission ends when you delete the material or its retention period expires (we then stop any new processing and transfer; for data already sent to upstream providers, see section 4 of the Privacy Policy), except for records we are required by law to keep.
Model output may resemble other people's work and may not be protectable by copyright in every jurisdiction. Before you use, publish or commercially exploit a deliverable, you are responsible for confirming that it complies with applicable law and third-party rights; use of platform characters and looks is also governed by the Creator Agreement.
8. Acceptable use
You must not use the Service to break the law, infringe the rights of others (including copyright, trademarks, likeness and privacy), pursue fraudulent or other harmful or unauthorised purposes, or produce content prohibited by the Content Policy.
You must not interfere with or disrupt the Service, including by bypassing sign-in, key permissions, rate limits, concurrency limits or spending caps; creating accounts in bulk by automated means; probing or attacking our systems; or reverse engineering the Service beyond what the law allows.
You may integrate the Service into your own products or workflows and deliver the outputs to your clients. Without our written consent, however, you must not give your API keys to third parties, or resell, sublicense or repackage access to the Service itself as another generation API for others. Any written consent from us will state the models, recipients and purposes it covers and will not exceed the authorisation our upstream providers give us; delivering outputs to clients does not give you a right to resell the API.
You must not claim that your product is endorsed by, partnered with or approved by us unless we agree in writing.
9. Changes, suspension and termination
We keep developing the Service and may add, change or discontinue features, models, parameters or prices; confirmed quotes are handled on the terms in effect when they were confirmed.
We may suspend or terminate your account or a particular key if you breach these Terms or the Content Policy; if there is reasonable evidence that you raised a chargeback knowing the reason was untrue, or that fraud is involved (a payment dispute raised for a valid reason is not in itself a breach; see section 3 of the Refund Policy); if your use creates a security risk for the Service or other users; or if the law, an authority or an upstream provider requires it. Where circumstances allow, we will notify you in advance and explain why; for a breach that is not urgent and can be remedied, we will first give you a reasonable period to remedy it. You may object to a suspension or termination by writing to support@avidol.ai.
You may stop using the Service and revoke your keys at any time. If we terminate your account for a reason other than your breach of these Terms (including because we stop providing the Service), you may write to support@avidol.ai to request a return of your unused top-up principal credits (excluding granted and bonus credits), which we handle manually as described in section 3 of the Refund Policy; held credits are settled under section 4 first. If we terminate because you breached these Terms, we may deduct from your unused top-up principal credits any amounts you owe us and the verified losses and costs your breach caused us, with an itemised statement; the remainder is returned to the extent applicable law requires.
10. Intellectual property
The software, API, documentation and interface design of the Service, and the avidol name and logo, belong to us or our licensors. Except as these Terms expressly allow, you may not copy, modify, distribute or sell them.
If you choose to send us suggestions or feedback, we may use them freely to improve the Service without payment or attribution.
11. Disclaimer
The Service is provided as is and as available. Output is produced by models and may not match what you expected, may contain errors, and may resemble other people's work. To the extent the law allows, we do not warrant that the Service will be uninterrupted, error-free or fit for a particular purpose.
12. Limitation of liability
To the extent the law allows, we are not liable for indirect, incidental, special, consequential or punitive damages, or for loss of profits, revenue or goodwill. If data is lost, our liability depends on our degree of fault and applicable law; please download and back up the files you need within their retention period.
To the extent the law allows, our total liability to you in connection with the Service is capped at the amount you actually paid us in the 12 months before the event giving rise to the claim. Amounts we must return to you under these Terms or the law (including unused top-up principal) do not count towards this cap.
The two limits above do not apply to our wilful misconduct or gross negligence, or to any liability that cannot be excluded or limited under applicable law; if any part of this section is unenforceable, the rest remains in effect.
If you use the Service as a consumer, the first two paragraphs apply only to the extent applicable consumer protection law allows.
13. Indemnity
If you use the Service for business or professional purposes and a third party brings a claim against us arising from content you provide, your use of a deliverable, or your breach of these Terms or the law, you will compensate us for the resulting losses and reasonable costs (including legal fees), excluding any damage attributable to us. We will notify you once we become aware of the claim and let you take part in handling it, and we will not settle in a way that increases your obligations without your consent. This section does not apply if you use the Service as a consumer.
14. Tax
We currently do not withhold any tax. We currently provide receipts for top-ups; general billing documents and invoices that qualify under local tax law (such as Japan's qualified invoices) are not currently provided, and if we provide them in future we will explain separately; the two are not the same. The amount shown on the checkout page is the total you pay us for that top-up; we will not add tax or any other undisclosed charge after payment (fees charged by your card issuer or payment service are separate). Whether and how tax applies is determined by applicable law, which may include the law of where you are; the details will be announced after review by a tax adviser.
15. Governing law and disputes
If a dispute arises, please first write to support@avidol.ai; both sides will try in good faith to resolve it by negotiation.
These Terms are governed by the laws of Hong Kong, and disputes arising from these Terms or the Service are subject to the jurisdiction of the courts of Hong Kong.
16. Changes to these Terms and contact
Updates are published on this page with a new version and effective date; this version takes effect on 11 November 2026. A change to these Terms does not ask you to tick the acceptance box again. Questions: support@avidol.ai.
For material changes that are unfavourable to you, we will post the change on this page at least 14 days before it takes effect.
Restrictions added by these Terms do not apply retroactively to transactions made before they take effect (for example top-ups already credited or quotes already confirmed); your rights under the law are not affected.
These Terms and the policies are written in Traditional Chinese and English, and both versions have equal force; where they differ, they are read together in light of the purpose of the document as a whole, and when we find a difference we will correct it and mark the correction on this page. The Simplified Chinese and Japanese versions are translations. No language-priority rule reduces the rights you have under applicable mandatory law.
Changelog
- v2026-09 · 2026-11-11 · initial