Privacy
Last updated: October 21, 2026
What we receive
Only the file you upload, its hash (SHA-256), and an IP hash used to count daily usage.
The IP hash uses a value (salt) that changes every day; we do not store your IP address. There are no accounts.
How it is handled
The file is sent to the analysis server and deleted as soon as analysis finishes — within one hour at the latest, with a separate daily cleanup that removes any leftover files.
No person looks at your file, and it is not used to train models. We do not identify people by face or voice, and we do not create or store identifying features.
What we keep
The result (label and scores) and the file hash for 30 days, then they are deleted. The IP hash is used only to count daily usage and is deleted after 2 days. The original file and preview images are not kept. If you report an error, your choice and note are kept with the result.
Transfers abroad
We use the providers below to run the service; some are outside Korea. Data is sent over encrypted connections when you use the service. If you do not want this transfer, please do not use the service; you will not be able to get an analysis.
| Recipient (location) | Data | Purpose | Retention |
|---|---|---|---|
| Vercel Inc. (USA, runs in Seoul) | Request data (IP address, file hash) | Running the website, managing jobs | While handling the request (logs per provider default) |
| Modal Labs, Inc. (USA) | Uploaded file | Analysis | Only during analysis |
| Supabase Inc. (Seoul region) | Uploaded file, result, file hash, IP hash | Holding the file before analysis, showing results, usage limits | File deleted right after analysis (within 1 hour); result and file hash 30 days; IP hash 2 days |
Age
The service is not available to children under 14.
Contact
Privacy inquiries: {contact}
Sample file sources and licenses
These files are the samples on the home page. Credits and changes are listed below.
| File | Credit · source | License | Changes |
|---|---|---|---|
| ai_photo.jpgAI-generated photo (generator: Absolute_Reality) | Zhenglin Huang et al. So-Fake-Set (Hugging Face, training row 40) | CC BY 4.0 | JPEG re-encode served by the Hugging Face dataset viewer (re-encoded from the original) |
| real_photo.pngReal photo ("chelsea") | Stefan van der Walt scikit-image data | CC0 | None |
| ai_voice.wavKorean synthesized speech (model: Fun-CosyVoice3-0.5B-2512, using the official zero_shot_prompt example voice; not a clone of any dataset speaker) | FunAudioLLM Fun-CosyVoice3-0.5B-2512 (Hugging Face) | Apache-2.0 (model) | Text synthesized into speech |
| ai_voice.wavText used for synthesis (line 3) | Zeroth Project; Lucas Jo (Atlas Guide Inc.), Wonkyum Lee (Gridspace Inc.) Zeroth-Korean (Hugging Face copy; original release OpenSLR 40) | CC BY 4.0 (text) | Only the text is used, synthesized into speech |
