1. What the service is
iLoveLTX is a compute provider. You choose a public weight (an LTX base model or a LoRA listed in the catalog or brought by you) and a workflow; we run that pipeline on our GPUs for your account and give you the output file. We do not choose the weights, we do not edit your prompt, and we do not keep a copy of the output beyond what is needed to deliver it. iLoveLTX provides compute, not output storage: a produced file is available for 24 hours at most, then deleted without backup.
The catalog aggregates public information about LTX LoRAs: where the weights are, the workflow that runs them, the traction they have, and what their author said about licensing. Downloads point to the author's own page or repository.
2. What you pay for
The unit of compute is the second of video produced, priced at the reference resolution and scaled by the number of pixels and by the speed you pick. A failed job is not billed. Downloads and submissions are free. Prices are shown on every tool page and in GET /v1/models before you run anything.
3. Labels are information, not decisions
Each catalog entry carries one of three labels. They summarize our reading of the author's declared permissions for a third-party generation service. They are not legal advice and they do not replace your own check.
The author published an explicit license compatible with running the weight on a third-party service. Nothing to do on your side beyond the license's own terms (credit, for instance).
The license is missing, conditional or ambiguous. All rights stay reserved by default: get the author's permission before you use the output commercially. The source page gives you the contact.
The author explicitly restricted third-party generation services, for example by allowing only the Civitai generator. You need the author's written permission before running the weight here.
Every entry we can run technically is runnable whatever its label. Before the first run of an entry that is not Cleared, you confirm once, for your key, that you have checked the license of that weight for your use and take responsibility for the output. We record that confirmation with the date, your key, your IP address and the catalog version you read.
4. Your responsibility
- You are responsible for the weights you choose: their license, the permissions their author granted or withheld, and the rights of the people and works they may reproduce.
- You are responsible for the output: what you generate, how you use it, and where you publish it. Content that impersonates a real person without consent, that reproduces a copyrighted character for commercial use, or that is illegal where you or your audience are, is your liability.
- You keep your API key private. Runs made with your key are yours.
5. Our responsibility
- We run what you asked, on the weights you named, and deliver the file, or we do not bill.
- We keep the labels current to the best of our knowledge and change them when an author tells us otherwise.
- The demonstration renders we publish ourselves use Cleared entries only.
- We do not offer adult content on our compute, and we do not serve full replacement models; those entries are listed for download only.
6. Authors and rights holders
If you made one of the LoRAs listed here, the claim page lets you get credit on the card, choose what we may do with your weight (run it here, list it for download only, or neither) and change its label. If you hold rights on a work used without permission, the same page takes a takedown notice: we acknowledge it at once, pull the links and the listing within 48 hours pending review, and notify the uploader, who may file a counter-notice within 10 business days. If they do not, the removal stands.
7. Outputs and storage
- iLoveLTX provides compute, not output storage. A produced file is made available for delivery, for 24 hours at most, then deleted without backup. Each file in the API response carries its
expires_at. - It is your responsibility to download the file within that window. After it, the download link answers that the file is gone.
- No copy is kept and no output is reused by the platform. Demonstration renders are produced separately, on Cleared entries.
- Job metadata (id, parameters, price, duration, timestamps) is kept for billing and support, without the file.
- Paid storage of custom workflows and LoRAs is a separate service, billed by weight, and never covers outputs.
8. Custom workflows and LoRAs
- When you order a workflow or deposit a private LoRA, we store its files on our compute so that your pipeline can run. That storage is billed monthly, by the real weight of the files, at the rate shown on the Build page and in
GET /v1/public-config(credits per GB per month, 1 credit = $0.01), charged to your balance: a share of the month when the asset goes live, then a full month on the first of each month. You accept this fee explicitly when you order; the rate you accepted is kept with your consent. - You can pause an asset at any time from your account: it is kept at the reduced rate and cannot run. You can delete it at any time: its files are erased from our compute and from our storage, and billing stops the same day; the current month is not refunded.
- If your balance cannot cover the fee, we tell you and give you a grace period (7 days by default). Then the asset is paused automatically. An asset left paused for a long time (90 days by default) is deleted, after a notice one week before.
- Private workflows run by API and MCP only, with the keys of the account that owns them. They never appear on the site. LoRAs you contribute to the public catalog, once accepted and listed, are not billed to you.
- This storage never covers the outputs of your jobs, which follow section 7.
9. Adult content
- Adult entries of the catalog run by API and MCP only, with your key. They are never offered in the site preview, and we never use them for our own demonstration renders.
- Before the first run you confirm, once per key, that you are an adult, that adult content is legal where you are and for your audience, and that you take responsibility for the output. That confirmation is recorded like the other consents.
- Real people without documented consent and minors are excluded whatever the label of the weight. Reports are handled through the claim page; an entry is pulled within 48 hours pending review.
- Outputs follow section 7: made available for 24 hours, then deleted.
10. Catalog sources
- The labels of the catalog reflect what the author declared on the source page (Civitai, Hugging Face) as recorded on a given date. iLoveLTX keeps a dated archive of that page, its API data and a screenshot as proof of that declaration. The archive is internal, never republished, and produced on request in case of a dispute.
- Source pages are checked regularly. When a page has moved, the entry follows it. When it is gone, the entry says so, its source link is struck through, its label is frozen at the last recorded declaration, and you see that before you choose to run it. Authors and rights holders can update or remove the entry from the claim page.
- We do not republish the examples of a gone page, and we do not mirror the weights of entries we do not serve.
11. Changes
These terms carry a version date. When the text changes, the version changes and every key accepts the new version before its next run. The version you accepted is kept with your consent record.