Privacy Policy
Effective date: November 9, 2026
Last updated: November 9, 2026
What changed (November 9, 2026): we now say that the FLUX.2 Pro image model, made by Black Forest Labs, runs through fal.ai, and that fal.ai passes your prompt, parameters, and any reference images to Black Forest Labs, which generates the image and may do so outside the United States (Sections 3 and 4). The model has always worked this way; what we do with your data has not changed.
What changed (November 9, 2026): the canvas now has image tools for background removal and upscaling. When you use them, the image you process is sent to fal.ai, which downloads it from our storage; for background removal, fal.ai also passes the image to Pixelcut, which makes the model that removes the background and may process the image outside the United States (Sections 3 and 4).
What changed (November 9, 2026): before an image is generated, we now send its full prompt to OpenAI's moderation service to check it against our Acceptable Use Policy, and before a generated image is saved, we send the image as well (Sections 2 and 3).
What changed (November 9, 2026): we corrected how we describe our AI providers' handling of your content; what we do with your data has not changed. We no longer say that fal.ai does not use the content we send it to train its models, because fal.ai does not publish that commitment. Instead we describe what we can confirm: we ask fal.ai not to keep a record of our requests, fal.ai hosts generated image files for a period set by its own rules, and we copy them to our own storage (Sections 3 and 5).
What changed (November 9, 2026): we added the Recraft V4 Vector model, made by Recraft, which creates SVG vector images. It runs through fal.ai, which passes your prompt and parameters to Recraft to generate the image; Recraft may process them outside the United States (Sections 3 and 4).
What changed (November 9, 2026): you can now delete your account yourself in Settings. Deletion takes effect immediately; requests by email are still handled within 30 days (Sections 5 and 6).
What changed (October 30, 2026): we added the Seedream 4.5 image model, made by ByteDance, and the Ideogram V3 image model, made by Ideogram. Both run through fal.ai, which passes your prompt, parameters, and any reference images to that company to generate the image; they may process them outside the United States (Sections 3 and 4).
What changed (October 25, 2026): our contact email is now [email protected]; nothing else changed.
What changed (October 24, 2026): we now say that the Nano Banana 2 and Nano Banana Pro image models, made by Google, run through fal.ai, and that fal.ai passes your prompt, parameters, and any reference images to Google to generate the image (Section 3). What we do with your data has not changed.
What changed (October 21, 2026): we now use Sentry to receive technical error reports when a page or request fails; these reports are set up not to keep your IP address, cookies, account details, or anything you typed. Reports from the website are sent from your browser directly to Sentry (Sections 1.5, 2, 3, and 5).
What changed (October 20, 2026): we corrected and completed some descriptions; what we do with your data has not changed. We now list more of the services your browser connects to directly, and describe the short-lived counters we keep to prevent abuse (Sections 1.5 and 5). We also explain that fal.ai receives the reference images you attach and may serve generated images to your browser (Section 3), correct the list of purchase details we keep (Section 1.4), and say what you can change in account settings (Section 6).
What changed (October 18, 2026): paid plans and credit packs are sold through our reseller Paddle. We describe what purchase data Paddle and we each handle, and that Paddle is an independent controller of the payment data you give it.
What changed (October 13, 2026): our website analytics now also counts a few named actions, such as clicking a sign-up button; these events contain no personal details you typed.
What changed (October 13, 2026): website analytics data is now kept for as long as we need it, rather than for a fixed 12 months.
What changed (October 12, 2026): we now describe the cookieless website analytics (Umami) we use, and list the short-lived
invite-codesign-up cookie.
This Privacy Policy explains how Far Star Studio Ltd. (遠星工作室有限公司) ("Far Star Studio", "we", "us"), a company incorporated in Taiwan (Republic of China), collects, uses, and protects personal data when you use Content Forger (the "Service") or join our waitlist. We are the data controller for the personal data described here.
Contact for all privacy matters: [email protected]
This policy is written to comply with Taiwan's Personal Data Protection Act (PDPA) and, where applicable to you, the EU/UK General Data Protection Regulation (GDPR). We aim to describe only what the Service actually does today; when the Service changes, this policy will be updated.
1. Data We Collect
1.1 Account data
- Email address (required)
- Full name (required at sign-up)
- Password — stored only as a secure hash by our authentication provider (Supabase); we never see your plaintext password
- Profile picture and account identifiers from Google or GitHub, if you choose to sign in with those providers
1.2 Waitlist and invitations
If you join the waitlist or redeem an invitation code, we collect:
- Your email address
- Your answers to the waitlist questions (free-form survey responses)
- How you found us (referral source) and, for invitation codes, the campaign the code belongs to
We use this to manage early access and understand who is interested in the Service.
1.3 Content and generation history
- Your content: projects (including project context/guidelines), workflows, entry set data, and uploaded assets
- Generation history: the Service stores the complete inputs and outputs of each generation step — including your prompts, the project context sent with them, and the full generated text or images — so that you can review past runs
1.4 Usage, purchases, and billing
- Credit balance and a ledger of credit transactions (when credits were granted, spent, returned, or expired, and on what)
- Purchases. Paid plans and credit packs are sold by our reseller Paddle (Section 3.1), which collects your payment details, billing name and address, and tax information directly. We never receive or store your card or bank details. When you buy something, we receive from Paddle and keep: Paddle's identifiers for you, your subscription, and your transactions; which plan or credit pack you bought, its price and currency, and when; and the subscription's status and billing dates.
1.5 Technical data
- Website analytics: we use Umami, a self-hosted, cookieless analytics tool, to measure page views and a few named actions on the website — for example, clicking a sign-up button, completing sign-up, or joining the waitlist. An action event contains only the action's name and where on the page it happened; it never includes your email address, anything you typed, or your content. For each page view or action it receives the page address (including any query string, except that invitation codes are removed and internal identifiers such as project IDs are replaced with a placeholder), page title, referring page (with the same parts removed), screen size, and browser language, plus your IP address and browser user-agent as part of the request. From these it derives your approximate location (country, region, city), your browser, operating system, and device type, and a visitor identifier that is reset every month. It does not set cookies or store your IP address, and it is not linked to your account. See the Cookie Policy for how to opt out.
- Error reports: when a page or request fails, the website or our servers send a technical error report to Sentry. A report contains the error message and where in our code it happened (stack trace), the page address without its query string, the type and version of your browser and operating system, and a short trail of the technical events just before the error on that page (such as page navigations, clicks on page elements, and failed network requests, with addresses again stripped of query strings). Sentry receives your IP address as part of the request your browser makes, but we configure it not to store it. We also configure error reports not to include cookies, request headers or contents, your account details, or what you typed into the Service, and we do not record your screen (no session replay). Error reports are not linked to your account and are not used for analytics or advertising.
- We do not run any advertising trackers or third-party analytics services.
- Our hosting and infrastructure providers keep standard server logs (IP address, browser user-agent, request timestamps) for security and operations
- Your browser also connects directly to some services, which therefore see your IP address and browser user-agent as part of serving those requests: our storage provider (Cloudflare), when you upload or view assets; our authentication provider (Supabase), when you sign in with Google or GitHub and when a workflow run shows its progress live; Google or GitHub, which serve your profile picture if you signed in with them; fal.ai, when a generated image is shown from fal.ai's address (Section 3); and our error-monitoring provider (Sentry), which receives error reports when a page or request fails
- Abuse-prevention counters: to limit abuse such as too many sign-in, sign-up, password-reset, or waitlist attempts, we count recent attempts per email address and per IP address. Each counter is stored under a hash of the email address or IP address instead of the address itself. Such a hash can still be matched to a known email address or IP address, so we treat it as personal data. Counters for IP addresses are kept only in our server's memory; counters for email addresses are kept in our database. Each counter covers at most one hour of attempts (see Section 5 for when it is deleted).
2. How We Use Data
We use personal data to:
- Provide and operate the Service (accounts, projects, generation, credit accounting)
- Process your purchases with Paddle: add the credits you paid for, keep your plan status up to date, handle cancellations, refunds, and billing questions, and keep the records the law requires
- Measure aggregate website usage (page views, a few named actions such as sign-up clicks, referring sites, device types) to improve the Service
- Send transactional emails through our authentication provider (e.g., sign-up confirmation, password reset). We do not currently send marketing emails.
- Manage waitlist access and invitations
- Find and fix errors in the Service (error reports, Section 1.5)
- Check image prompts and generated images against our Acceptable Use Policy with an automated moderation service (Section 3)
- Secure the Service, prevent abuse and fraud, and enforce our Terms of Service
- Comply with legal obligations
We do not sell personal data, use it for advertising, or use your content to train AI models.
For users in the EEA/UK, our legal bases are: performance of a contract (operating the Service for you), legitimate interests (security, abuse prevention, service improvement, error monitoring, and cookieless website analytics), consent (waitlist participation), and compliance with legal obligations.
3. Who We Share Data With
We share data only with the processors needed to run the Service, and with Paddle when you buy something (Section 3.1):
| Provider | What they process | Location |
|---|---|---|
| Supabase | Authentication (email, password hash, OAuth identities) and our database (all data in Sections 1.1–1.4) | United States |
| Cloudflare | Asset storage — files you upload and images you generate; sees your IP when your browser uploads/downloads assets | United States (global network) |
| OpenAI | Text generation — receives your prompts and the project context attached to them. Content moderation — receives the full prompt of each image generation (including any style, project context, and system prompt added to it) and each generated image (OpenAI downloads it from fal.ai), to check them against our Acceptable Use Policy | United States |
| fal.ai | Image generation — receives your image prompts and parameters, and any reference images you attach (fal.ai downloads them from our storage). Canvas image tools (background removal and upscaling) — receives the images you process with them, which fal.ai also downloads from our storage. Generated images are first hosted by fal.ai; if copying one to our storage fails, the app shows it from fal.ai's address, so your browser loads it from fal.ai. Note: the "GPT Image 2" image model, made by OpenAI, is also routed through fal.ai. So are the "Nano Banana 2" and "Nano Banana Pro" image models, made by Google: when you use them, fal.ai passes your prompt, parameters, and any reference images to Google, which generates the image. The same goes for the "Seedream 4.5" image model, made by ByteDance, the "Ideogram V3" image model, made by Ideogram, the "Recraft V4 Vector" image model, made by Recraft, and the "FLUX.2 Pro" image model, made by Black Forest Labs: fal.ai passes your prompt, parameters, and any reference images to that company, which generates the image and may do so outside the United States (Section 4). Background removal uses a model made by Pixelcut: when you use it, fal.ai passes the image to Pixelcut, which removes the background and may do so outside the United States (Section 4) | United States |
| Railway | Application hosting and server logs, including our self-hosted analytics (Umami) | United States |
| Sentry | Error monitoring — technical error reports from the website and our servers (error message, page address, browser type; see Section 1.5) | United States |
| Google / GitHub | Sign-in, only if you choose OAuth login; they process your login per their own privacy policies | United States |
How our AI providers handle the content we send them:
- OpenAI (text generation and content moderation): OpenAI's API data policy states that content sent to its API is not used to train OpenAI's models unless the customer opts in to share it.
- fal.ai (image generation, for every image model in the app, and the canvas image tools): we ask fal.ai not to keep a record of our requests (the prompts, parameters, reference-image links, and output links it would otherwise keep in its request history). This request covers fal.ai's own records only; for the Nano Banana, Seedream 4.5, Ideogram V3, Recraft V4 Vector, and FLUX.2 Pro models, fal.ai also passes the content to Google, ByteDance, Ideogram, Recraft, or Black Forest Labs respectively, and for background removal it passes the image to Pixelcut, as described in the table above. It also does not cover the generated image files: fal.ai hosts those for a period set by its own rules (for background removal, Pixelcut may host the result instead, under its own rules), and we copy each one to our own storage right after it is generated (see the table above for what happens if that copy fails). How fal.ai otherwise uses the content it receives is governed by its Terms of Service and Privacy Policy.
3.1 Paddle (our reseller)
Paid plans and credit packs are sold by Paddle, our online reseller and Merchant of Record (see our Terms of Service). Depending on where you are, the seller is Paddle.com Market Limited (United Kingdom), Paddle.com Inc. (United States), or Paddle.com (Canada) Ltd.
Paddle is not one of our processors. When you check out, you give your payment and billing details directly to Paddle, which uses them to take payment, calculate and pay taxes, prevent fraud, and provide receipts and buyer support, as described in Paddle's Privacy Policy. Paddle and we each decide independently how we use the purchase data we hold, and each of us is responsible for its own use: we are independent controllers.
We share with Paddle only what it needs to connect a purchase to your workspace: an internal reference for your workspace, the plan or credit pack you chose, and, to fill in the checkout form for you, your account email address. Paddle shares back the purchase details listed in Section 1.4.
We may also disclose data where required by law, court order, or to protect the rights and safety of users or the public (including mandatory reporting of child sexual abuse material).
If the Company is involved in a merger, acquisition, or asset sale, personal data may be transferred as part of that transaction; we will notify you of any such change.
4. International Transfers
We are a Taiwan company, and our infrastructure currently stores and processes data in the United States through the providers listed above. The exceptions are the image models made by ByteDance, Ideogram, Recraft, and Black Forest Labs and the background-removal model made by Pixelcut (Section 3): fal.ai passes your requests for them to those companies, which may process them outside the United States. We may additionally use infrastructure in Singapore in the future; if we do, we will update this policy. Purchases are handled by Paddle in the United Kingdom, the United States, or Canada, depending on where you are (Section 3.1).
Where GDPR applies, transfers to these providers rely on appropriate safeguards such as the providers' standard contractual clauses and data processing agreements. By using the Service you acknowledge that your data is processed in the locations described above.
5. Retention
-
Account, content, and generation history: retained while your account is active. Where the app offers deletion (e.g., deleting an asset or a workflow), deleted content is removed from the live database and storage.
-
Waitlist data: retained while we operate the waitlist, so we can grant access later.
-
After account deletion (Section 6): when you delete your account in Settings, your account, workspace and content are deleted from our live database immediately, and the files you stored are removed from storage shortly afterwards. When you ask us to delete your account by email, we do this within 30 days of verifying your request. Residual copies in encrypted backups are purged on our providers' backup rotation schedules.
-
Website analytics data: retained for as long as we need it to understand how the website is used. It does not include your IP address and is not linked to your account (Section 1.5).
-
Error reports: deleted automatically by Sentry after 30 days.
-
Server logs: retained on our infrastructure providers' standard short-term schedules.
-
Generated image files at fal.ai: hosted by fal.ai for a period set by its own rules; the copy we keep in our storage follows the rules for your content above (Section 3).
-
Abuse-prevention counters for email and IP addresses (Section 1.5): each counter covers at most one hour of attempts. After that it is no longer used for anything and is deleted automatically: from server memory when the server restarts or needs the space, and from our database by a background cleanup a day or more after the counter expires.
-
Purchase records (Section 1.4): kept while your account exists and afterwards for as long as tax and accounting law requires.
-
We may retain limited data longer where required by law or needed to resolve disputes or enforce agreements.
6. Your Rights
Subject to applicable law (including the Taiwan PDPA and, where applicable, GDPR), you have the right to:
- Access the personal data we hold about you
- Correct inaccurate data (you can change your email address and password in account settings, except that the email address of an account that signs in only with Google or GitHub comes from that provider; for anything else, email us)
- Delete your account and personal data
- Export your data in a machine-readable format (portability)
- Object to or restrict certain processing, and withdraw consent where processing is based on consent
How to exercise these rights: you can delete your account yourself in Settings → Account → Delete account. Deletion takes effect immediately and cannot be undone. For your other rights, including data export, or if you cannot sign in, email [email protected] from the email address registered to your account (we verify requests this way to protect your data). We handle these requests manually and respond within 30 days.
Deleting your account deletes the purchase records we hold, except those we must keep by law (Section 5). It does not delete the data Paddle holds as an independent controller; to exercise your rights over that data, contact Paddle as described in its privacy policy. You cannot delete your account while a subscription is active; cancel it first and wait until the current billing period ends.
If you believe we have not handled your data lawfully, you may lodge a complaint with your local data protection authority (for EEA/UK users) or the competent Taiwan authority.
Note for California residents: we do not sell or share personal information for cross-context behavioral advertising.
7. Security
We protect data using encryption in transit (TLS), passwords stored only as secure hashes, scoped access credentials for infrastructure, and application-level authorization that restricts every request to the requesting user's own workspace, with row-level security policies additionally defined in the database. No online service can guarantee absolute security; we recommend using a strong, unique password.
If a data breach occurs that is likely to result in a risk to your rights, we will notify affected users and the competent authorities as required by applicable law, without undue delay.
8. Children
The Service is not directed to anyone under 18, and we do not knowingly collect data from them. If you believe a minor has provided us data, contact [email protected] and we will delete it.
9. Cookies
We use only essential and functional cookies — no analytics or advertising cookies. Our website analytics (Umami) is cookieless. When you open checkout, Paddle runs its own checkout, which may set cookies needed for payment and fraud prevention. See the Cookie Policy for details.
10. Changes to This Policy
We will update this policy as the Service evolves (for example, when our analytics setup changes). Material changes will be announced on the website and the date at the top updated. The English version of this policy governs; translations, if provided, are for reference only.
11. Contact
Far Star Studio Ltd. (遠星工作室有限公司)
Taiwan (R.O.C.)
Email: [email protected]