Privacy Policy

This page describes what Lucy Labs ("we," "us"), part of the Astryks Group, collects and does with your data when you use lucylabs.app and the Lucy Labs mobile app. It's written in plain language and reflects what we actually do — not boilerplate we haven't checked against the real product.

Text you type for narration, any voice recording you upload to clone, and the audio we generate are processed entirely on our own infrastructure — GPU servers we control, rented from RunPod and Modal. We do not send this data to any third-party AI vendor. Generated audio is not stored on our servers — it's returned directly to your device for playback and download, so save anything you want to keep. Uploaded voice samples are used only to generate your requested audio and are not retained afterward or used to train models for other users.

Video generation is a live, paid feature (talking-head avatars, cinematic scenes, and our pay-as-you-go engine picker). It works differently from audio-only narration/voice cloning above, and we want to be direct about that difference rather than bury it:

Any photo, video, or reference audio you upload for a video generation — along with the text/audio we generate on your behalf to drive it — is sent to third-party AI model providers to actually produce the video. Depending on which engine you (or the feature) select, that means one or more of: Kling (Kuaishou), Veo (Google), and Seedance (ByteDance) — all reached through fal.ai, the inference platform that hosts and routes to these models on our behalf. fal.ai also runs a small audio/video-merge utility we use when a video needs your uploaded or generated audio track attached to it.

Practically, this means: your uploaded photo/video/audio and the resulting generated video are transmitted to and briefly stored by fal.ai and whichever model provider actually renders it, are subject to those companies' own privacy and retention policies (not just ours), and the finished video is returned to you via a URL hosted on fal.ai's infrastructure rather than ours. We do not control how long fal.ai or its model providers retain that content on their own systems, and we have asked fal.ai directly about aspects of this we could not confirm from public documentation alone. We never send your data to these providers for anything other than fulfilling the specific video you asked for — not for their model training, and not for ours.

We don't use passwords or traditional accounts. Instead, a subscription is tied to an access code generated after checkout. Payment is processed by Stripe — we do not see or store your card details ourselves. We store your email, subscription status, and usage (characters/video credits used this billing period) to enforce plan limits and manage your subscription. Your email address and subscription plan name are also sent to Resend, the service we use to actually deliver your access-code, sign-in, and payment-failed emails.

Summarizing the companies above in one place, and what each one actually receives from us:

• fal.ai, Kling, Veo, Seedance — your uploaded photo/video/audio and generated video, only when you use a video feature (see "Video" above).
• Stripe — payment details and billing email, to process your subscription or credit purchase. We never see your raw card number.
• Resend — your email address and plan name, to deliver transactional emails (access codes, sign-in links, payment-failure notices).
• RunPod, Modal — audio you submit for narration/voice cloning, and photo/video/audio for video generation before it's forwarded to the vendors above. These are GPU servers we rent and run our own code on, not third-party AI services that process data on their own terms.
• Vercel — hosts the site and, for signed-in users, your generation history (see "Data retention & deletion" below).

We keep a lightweight, anonymous record of page visits (a randomly generated id stored in your browser, the page path, and a timestamp) so we can see roughly how much traffic the site is getting. We don't collect your IP address, device fingerprint, or any identifying information as part of this, and it isn't linked to your account or billing data.

• We don't sell your data.
• We don't use your uploaded voice, photos, or video to train models — not for anyone else's benefit, and not for ours.
• We don't send audio-only narration/voice-cloning data to any third party — that stays entirely on infrastructure we control (see "Audio" above). Video generation is the one exception, disclosed plainly above, not buried here.

If you generate audio without signing in, nothing is stored on our servers — it exists only in your browser for that session. If you sign in, we keep a short history of your generations (currently 14 days) so you can play them back and re-download them from your account; anything older is automatically deleted. Videos are not stored on our own servers at any point — you get a direct link to the file as hosted by fal.ai, and signed-in users' history keeps that link, not a copy of the video itself, for the same retention window. If you'd like anything deleted sooner — your account, uploaded samples, or generated clips — email us and we'll take care of it. We can't force an early deletion on fal.ai's or a model provider's own systems, but we can and will ask on your behalf.

Questions about this policy or your data: support@astryks.com.