Privacy Policy

Loopcut is an open-source Blender application on your computer. With your ChatGPT plan, own API key or a local model, model requests go straight from your computer to the model, never through our server. The app fetches our service list, sends Coming soon clicks when you click those buttons, and sends a few usage events (app starts, setup steps and how each turn ended, never what you asked or your scene) unless you turn them off. From version 0.2.6 it also sends us a log of each conversation, on every model path, unless you turn sharing off: we use it to help you and to find and fix problems, never to train AI models, and delete it after 90 days. It never includes images, .blend files or your keys and tokens.

If you create a Loopcut account and use our hosted models, we hold the minimum needed to sign you in, meter your usage and bill you. Our gateway relays your model requests to the provider and does not store their content; the conversation log the app sends is the only copy we keep, and only while sharing is on.

Paid cloud tools (images, 3D, voice, sound effects and cloud render) are different: they need your files. We send each job only to the providers that run it, keep its files in our storage for 7 days and its results for 30 days after their last use, then delete them. Private mode, on by default for paid plans, keeps jobs away from providers that may train on them.

We also keep marketing counts for downloads, signups, first hosted requests and upgrades, two onboarding milestones and, while you are signed in, which model source the app uses, and count visits to this website without cookies. Signed-in Coming soon clicks are linked to your account; anonymous clicks use a random install id. These records carry no scene content.

Loopcut operates the website at loopcut.org and the hosted model service behind it (the “Service”). You can reach us at hello@loopcut.org.

Model requests do not pass through our server when you use your ChatGPT plan, own API key or a local model. On all three paths, the app still contacts loopcut.org in the three small ways below, and from version 0.2.6 it sends conversation logs as described under Conversation logs and problem reports.

It fetches the list of Loopcut services at startup and every 10 minutes while the panel is open. This needs no account and carries the app and Blender version.

If you click a Coming soon button, the app sends which service, where in the app you clicked, the app and Blender version, and a random install id the app generated. That id is not tied to an account or person. For anonymous clicks, the IP address is used only in hashed form to limit abuse. That abuse-limit record is deleted within about a day; the click itself is kept for 12 months. Signed-in clicks are linked to your account, as described under Measuring our marketing.

Usage events, unless you turn them off under Preferences › Add-ons › Loopcut. When the app starts, it sends the app and Blender version, the operating system (macOS, Windows or Linux), the processor type, the language Blender’s interface is set to and whether this is the first launch. During the setup window, it sends which step you reached, which option you picked (ChatGPT plan, Loopcut account, own key or skip), whether it worked, and a short error code if it did not. After each turn with the agent, it sends which kind of model ran it (ChatGPT plan, Loopcut or your own key), the model's name, whether the turn finished, failed, was cancelled or came back empty, how long it took, how many tool calls it made, and a short error code if it failed. Each event carries the same random install id as Coming soon clicks and, while you are signed in, your account. It never includes your prompts, the agent's replies, file names, paths or anything from your scene; the server rejects any field it does not expect. We use these events to find where setup and first turns go wrong. They are stored in our database for 12 months, and signed-in events are deleted with your account. As for clicks, no IP address or browser details are stored with them, and for anonymous events the IP address is used only in hashed form to limit abuse.

Apart from these service-list requests, Coming soon clicks, usage events, conversation logs and the problem reports you choose to send, the application does not send us analytics or crash reports.

On a phone, this website offers to email you the download link instead. We send that one email to the address you type, through Resend, and nothing else: you are not added to a list and get no follow-ups. We don’t store the address. To stop repeated requests, we keep keyed digests (HMAC) of the address and of your IP address, never either itself, and delete them within about two days. We count that a link was sent, with the page and the campaign tags or referring site of the link you followed, never the address. That count is kept for 12 months.

• Your email address, and if you sign in with Google, the name and profile picture Google shares with us. We use these to identify your account and, if you sign in by email, to send sign-in codes and links.
• A session cookie so you stay signed in on this website.
• The time zone of the connection you use Loopcut from, which our host works out from your IP address (we don't store the address), so account emails arrive at a sensible hour.
• When you connect the application, the device name it reports (for example your computer’s name), the app and Blender version and operating system it reports, when it last used the connection, and a hashed access token. We store the hash, not the token itself, so we cannot read it back. You can revoke any device from your account page.
• Email: when your account is created we send one welcome email with the download links and setup steps, and if you ask on the getting-started page we email you the download links. We record that each was sent. Every email has a one-click unsubscribe; after you use it we send nothing that isn’t about your account (sign-in emails and emails you ask for still arrive), and we keep the date you unsubscribed on your account.
• Billing records: a Stripe customer identifier, your plan and its status, and the current billing period. Card details are entered on Stripe’s pages and never reach our servers.
• That you confirmed you are 18 or older: when, and where (the sign-in page, connecting the app, your account page or checkout). Cloud tools need it once per account.
• Credit records: your monthly credit grants, top-up purchases, trial and promotional credits, and every hold, charge and return made by a cloud tool job.
• If your account receives the sign-up trial, a trial record: your account id and keyed digests (HMAC) of your email address and, when we have it, of the IP address you signed in from, never the address itself. It stops the trial being claimed twice, decides which free accounts get the free plan’s monthly credits, and is described under How long we keep it.
• Cloud tool jobs, if you use them: the records, files and results described under Paid cloud tools.
• While conversation sharing is on, the logs of your conversations, linked to your account while you are signed in, and any problem reports you send, as described under Conversation logs and problem reports.
• Your Private mode and no-generation settings, and any project overrides the app sends with a job.
• Marketing counts for signup and upgrades, and your first successful hosted request from the usage records. Coming soon clicks made while signed in are linked to your account, as are the model source the app reports and two onboarding milestones. Details and retention are below.

The first three paths send no model requests through our server. All three still fetch our service list and send Coming soon clicks as described above, and from version 0.2.6 the app sends a log of each conversation on every path unless you turn sharing off, as described next. Those logs never include viewport images.

When a signed-in user opens a survey or bug-report link from the app, the link carries a random account number, never the email, so founding-member credit can be matched to the account.

• From version 0.2.6, the app sends us a log of each conversation with the agent, on every model path (ChatGPT plan, own key, local model or Loopcut models) and with or without a Loopcut account. A log holds your prompts, the agent’s replies, its tool calls and their results, errors, the app’s log lines for that conversation, the app and Blender version, your operating system, which kind of model and which model ran it, and how each turn ended.
• A log never includes images (an image is replaced by a note of its size), .blend files, ChatGPT sign-in tokens or API keys. The app removes your keys and tokens before sending, and our server removes key-shaped text again before storing.
• We use logs to help you when something goes wrong, to find and fix problems and to improve Loopcut. We never use them to train AI models, and we do not sell them or share them with advertisers.
• Each log carries the app’s random install id and, while you are signed in, your account. It is stored in our private Cloudflare R2 storage, with an index of its versions, model and outcome in our database, and is deleted 90 days after the conversation’s last upload.
• Sharing is on by default. You can turn off “Share conversations to improve Loopcut” in the setup window or under Preferences › Add-ons › Loopcut. Turning it off stops future uploads; to delete logs already sent, email hello@loopcut.org.
• Report a problem, under Preferences › Add-ons › Loopcut, sends your note, a contact email if you give one, your app, Blender and system versions, the app’s recent log and up to three recent conversations, only when you press it, even with sharing off. We email its code, note, contact and versions to our support inbox and keep the report for 90 days.

When you are signed in and use Loopcut hosted models (Fast, Pro), our gateway relays each request to a provider in the model table and streams the reply back. Requests can include your messages, scene details the app chose to include and viewport images. This section covers the conversation with the agent; paid cloud tools are described below.

We record, per request: a request identifier, the model used, input and output token counts, the resulting cost, whether it succeeded, how long it took, and, when the app sends them, its conversation id and the agent role (for example orchestrator or worker). We use this to enforce your plan’s allowance and show your usage. The gateway does not store the content of your messages, your scene data, images or the model’s reply. The conversation log the app sends, described above, is kept separately and never contains images.

The model provider receives the request content in order to answer it. Their handling of that content is governed by their own terms and privacy policy.

Each hosted model is run by a model provider, and a busy model can pass a request to a backup. This table is generated from the gateway’s own routing table, so it lists every provider a hosted request can reach, whether that provider may use it to train models, how long it keeps it and where it processes it.

Cloud tools are jobs the app starts on your request and pays for with credits. For each job, the app uploads the files the tool needs to our private storage, and our server sends the tool’s input and those files to the one provider that runs it. Providers get files through short-lived signed links or in the request itself. They do not receive your Loopcut account details, sign-in tokens or the agent’s conversation. Finished results are copied to our storage and the app downloads them into your project.

Provider retention can include safety, abuse-monitoring and legal exceptions. Deleting a job’s files from our storage, or your account, cannot withdraw a licence a provider’s terms already gave it.

• Private mode is on by default for paid plans. While it is on, the server refuses any cloud tool whose provider may train on your data or has unclear terms; today that is sound effects. Tools whose providers do not train stay available, with the retention shown above. Private mode does not keep processing on your computer.
• The no-generation setting turns off tools that create new content: image generation and editing, image to 3D, voice lines and sound effects. It applies to your whole account, and can also be set for a single project. Tools that work on what you made, such as cloud render, remeshing and asset search, stay available and still use the providers above.
• Both settings are enforced on our server, not only hidden in the app. A refused job is not charged.

Our storage checks for expired files every hour and deletes them, so a file is normally gone within a day of its date. Account and billing records are kept while your account exists and for as long as we are required to keep financial records afterwards. Usage records for hosted models are kept while your account exists and are deleted with it. Revoked device tokens are kept only as revoked hashes so they cannot be reused. App usage events are kept for 12 months, or until you delete your account if you were signed in. Conversation logs are deleted 90 days after the conversation’s last upload and problem reports 90 days after you send them; deleting your account deletes those linked to it. Signed-in Coming soon votes, with the app’s random install id, stay until you take the vote back or delete your account; signed-out votes are deleted after 12 months. The trial record is kept after your account is deleted, with no end date, so the same email address cannot claim the trial again. Deleted data can remain in our hosting providers’ backups until those backups expire.

Paid plans are sold through Stripe Checkout and managed in Stripe’s billing portal. You enter card details on Stripe’s pages; we never receive the card number. From Stripe we keep your customer and subscription identifiers, the plan, whether it is billed monthly or yearly, its status and the current billing period. If your account has not yet confirmed that you are 18 or older, checkout asks; the confirmation is stored on your account, and a subscription also notes it in Stripe. Stripe processes payment information under its own privacy policy.

We count four things to learn which links bring people to Loopcut: downloads from this website, new accounts, an account's first successful request to our hosted models, and upgrades to a paid plan. For a download we record the time, the platform you chose, the release, and, if the link you followed carried them, its campaign tags (utm_source, utm_medium, utm_campaign, utm_content, utm_term) and the name of the site that linked to us (for example reddit.com, never the full address). A download is not linked to you or to an account. To stop automated downloads from inflating the counts, the IP address is used only in hashed form to limit how many downloads one address can add, and that record is deleted within about a day. When you create an account we record the same campaign tags and referring site, if your sign-in link carried them, and whether you signed up from the app or the website, linked to your account. When you upgrade we record the plan and the Stripe subscription identifier, linked to your account. The first hosted request is read from the usage records described above; nothing extra is stored. We do not record IP addresses, browser or device details, or any identifier that follows you from one visit to the next, and we set no cookies and use no browser storage for this: the campaign tags and referring site are held in tab memory while you browse, and added to the download and sign-in page addresses when needed. A reload or new tab does not preserve that memory. We use these records only in aggregate, to decide where to spend on marketing, and do not share them with advertisers. They are deleted after 12 months, or straight away when you delete your account.

When you are signed in to a Loopcut account, the app (version 0.2.4 or later) also tells our server which model source you use, ChatGPT plan, Loopcut models, your own key or a local model, each time it fetches the service list, and we store the latest value on your account. We also record two milestones linked to your account: when you have connected both a ChatGPT plan and a Loopcut account, and the first time you use a paid cloud tool. We use these to measure onboarding and conversion, and they carry no scene content. None of this is sent while you are signed out. The model source is replaced whenever it changes and deleted with your account; the milestones are deleted after 12 months, or straight away when you delete your account.

Separately, this website counts page visits with Vercel Web Analytics. For each page view Vercel receives the page’s path, the campaign tags and referring site described above if the link carried them, the site that linked to the page, your country, and your browser, operating system and device type. Any other part of the page address, such as a sign-in code, is removed in your browser before the count is sent. It sets no cookies and uses no browser storage: Vercel tells visits apart with a hash of the incoming request instead, and discards that hash after 24 hours, so a visitor cannot be followed from one day to the next. We see these counts only in aggregate.

When you are signed in and click a Coming soon button in the app, we record which service it was, where in the app you clicked, and the app and Blender version. This is linked to your account and helps us decide which paid services to build first. It carries no scene content. We delete it after 12 months or when you delete your account.

• Supabase, which hosts our database in the US and handles authentication, including email sign-in.
• Resend, which sends our emails (sign-in codes and links, the welcome email, the download links you ask for, and the problem reports you send, to our own inbox). It receives the addresses and the email’s content, in the US.
• Stripe, which processes payments and hosts the billing portal.
• Vercel, which hosts this website and the gateway, and counts visits to this website.
• Cloudflare, whose R2 storage holds cloud tool files and results, conversation logs and problem reports for the periods above.
• The providers that run paid cloud tools, each receiving only its own jobs: OpenAI, fal and the model developers it runs (Deemos for Rodin, Tripo), Google Cloud, ElevenLabs and RunPod (which runs our own render and motion workers, with NVIDIA’s GEM-X and Kimodo models inside them).
• Google, only if you choose to sign in with Google.
• Every model provider listed in the table above, including backup providers and OpenRouter as a relay, receives request content when a hosted request is sent that way. OpenAI receives requests sent straight from your computer under your ChatGPT plan; a provider you configure receives your own-key requests.

We do not sell personal data, and we do not share it with advertisers.

• You can use Loopcut without a Loopcut account, with your ChatGPT plan, own key or a local model.
• You can revoke any connected device from your account page at any time.
• You can turn off the app’s usage events under Preferences › Add-ons › Loopcut, and unsubscribe from email that isn’t about your account with the link in any email we send.
• You can turn off conversation sharing in the setup window or under Preferences › Add-ons › Loopcut, and ask hello@loopcut.org to delete logs already sent.
• You can cancel a paid plan from the billing portal linked on your account page.
• You can turn Private mode and the no-generation setting on or off in the app at any time.
• You can ask us to export or delete your account data by emailing hello@loopcut.org from the address on the account. We handle these requests by hand and confirm and complete them within 30 days. An export is a copy of your account, usage, credit and job records. Deletion covers your account, your job records, every file and result in our storage, and the conversation logs and problem reports linked to your account. We keep the financial records the law requires and the trial record above, and copies a provider already holds follow that provider’s retention above. Unused credits are not refunded.

Depending on where you live you may have additional rights under laws such as the GDPR or the CCPA, including the right to complain to a supervisory authority. We will honour them.

This website uses only the cookies needed to keep you signed in. There are no advertising or cross-site tracking cookies.

Loopcut account access tokens are stored hashed on our server, database access is restricted per user, secrets are kept out of the codebase, and traffic is encrypted in transit. Cloud tool files sit in private storage and are reached only through signed links. Upload links expire after 15 minutes and the links the app downloads results from after an hour. A link a provider uses to fetch a job’s files lasts as long as that job may run plus 5 minutes: a few minutes for quick tools, up to about 14 hours for the longest renders. ChatGPT sign-in tokens and your own API key are stored only on your computer, and the app removes them from conversation logs before sending; our server removes key-shaped text again before storing a log. The agent and the Blender build are open source under the GPL. This website and the hosted service are not open source.

Loopcut accounts are for people 18 or older, partly because the providers behind cloud tools require adult users, and every account is asked to confirm it once. We do not knowingly collect data from anyone under 18. If you believe a minor has created an account, contact us and we will delete it.

If this policy changes in a way that matters, we will update the date at the top and note the change in the project’s repository. Continued use of the Service after a change means you accept the updated policy.