Subprocessors

How to read this list

Each entry names the provider, what it does, and the data it receives. 'Only when triggered' means nothing is sent unless you actively use that feature; Privacy Mode disables the AI ones entirely (see /privacy-mode).

Does AetherCut store video content in MongoDB?

No. MongoDB stores account records and saved project metadata only when you sign in. It does not store video content. Stripe handles card and billing details directly when you upgrade to Pro.

• Emergent — application hosting and the AI gateway that routes our AI calls. Processes requests to the app and, for AI features, relays derived inputs to the model providers below.
• MongoDB — our database. Stores your account record and saved-project metadata (only if you sign in). No video content.
• Stripe — payment processing. Receives your card and billing details directly; we never see them. Only when you upgrade to Pro.
• Google (OAuth) — sign-in. Receives the standard OAuth handshake so you can log in with Google. Only if you choose to sign in.
• Resend — transactional email delivery (e.g. receipts, account notices). Receives your email address for messages we send you.

What data do AetherCut AI features send to providers?

AI tools send only the input needed for the feature you use, such as typed text, short audio clips, selected audio, prompts, or timeline metadata. Privacy Mode turns off the AI related subprocessors entirely.

• OpenAI — Whisper auto-captions (a few seconds of extracted audio, discarded after transcription), gpt-image-1 image generation (your text prompt), and TTS voiceover (your typed text).
• Anthropic (Claude) — AI Director, Voice Coach notes, Highlight Reel, and producer notes. Receives timeline metadata and short text, not video.
• Google (Gemini / 'Nano Banana') — image generation. Receives your text prompt.
• ElevenLabs — AI voiceover, voice enhance/clone, dubbing, and AI sound effects. Receives your typed text or the specific audio you choose to process.
• fal.ai — AI lip-sync, upscale, object eraser, and Kling text-to-video for AI B-Roll, AI Reshoot and Ghost Draft. Receives only the specific frames/clip you run the tool on, or your text prompt.

What does AetherCut share with media library providers?

AetherCut sends your search query to Pexels and Freesound when you search those libraries. If you import from YouTube, Google receives only the URL you paste.

• Pexels — royalty-free stock video/photo search. Receives your search query.
• Freesound — sound-effect and music search. Receives your search query.
• YouTube (Google) — used only if you import from a URL you provide. Receives the URL you paste.

Operational / SEO (no end-user data)

• Prerender.io — serves pre-rendered HTML to search crawler bots for our public marketing pages. Sees bot requests for public URLs only; no user data.
• Bing Webmaster Tools & Google Search Console — admin-only SEO dashboards. We send our own public site URLs to check index status; no user data.

Changes to this list

We update this page whenever a subprocessor is added or removed. To be notified in advance of changes for a contractual arrangement, email AethercutAdmin@gmail.com.

See also: /trust-security, /what-data-we-collect, and /what-data-we-do-not-collect.

Common subprocessor questions in practice

A simple way to think about these subprocessors is by action, not by vendor. If you trim clips, move items on the timeline, or export a project without using AI, that work stays on your device. If you sign in, MongoDB may store account details and saved-project metadata. If you pay, Stripe handles billing. If you search a stock library, your search terms go to that library. If you trigger an AI feature, only the specific input needed for that feature is sent to the relevant provider. That distinction usually answers the most common question: “what leaves my device, and when?”

A frequent misunderstanding is assuming that enabling the app means all providers can access a full project. That is not how this list works. The providers here are feature-specific, and several are used only after a deliberate action such as logging in, upgrading, searching Pexels, importing from YouTube, or running a tool like captions, dubbing, lip-sync, or image generation. In the same way, a prompt for image generation is not the same as uploading a clip, and timeline metadata for planning tools is not the same as sending raw video.

If you want the smallest possible data footprint, the practical approach is straightforward: stay signed out unless you need saved projects, avoid third-party library search unless you want external media, and keep Privacy Mode on when you do not want AI-related subprocessors involved at all. Before running any AI tool, check what you have selected. Choosing only the exact audio, clip, frames, or text you intend to process helps keep sharing limited to that narrow input rather than anything broader from the project.