1. The website
The current website is static. Its code does not create an account, accept form submissions, set first-party cookies, run advertising, or include a first-party analytics product.
This website is hosted on Cloudflare Pages. Cloudflare may process standard request and security information such as IP addresses, request headers, timestamps, and requested URLs. See the Cloudflare Privacy Policy.
Fonts are requested from Google Fonts and product images are requested from GitHub's raw-content service. Those providers receive the normal request information needed to deliver the files. See Google's Privacy Policy and the GitHub Privacy Statement.
2. Data stored by the desktop app
ClipsAnyone? stores its working data under
%LOCALAPPDATA%\ClipsAnyone. Depending on the features
you use, that can include source downloads, generated clips,
settings, transcripts, clip candidates, schedules and upload
history, feedback, learning data, game-information caches,
voice-profile data, and YouTube OAuth tokens.
Personalization records can include filenames or source identifiers, labels, timing, optional feedback reasons, and transcript excerpts. The learned aggregate preference model is stored locally and excludes the original media, paths, transcripts, identifiers, and voice embeddings.
Voice Profile is opt-in. It stores numeric acoustic characteristics and, when enabled, a numeric speaker-embedding profile rather than a copy of the enrollment audio. Temporary enrollment audio is removed after processing.
3. Network connections made by the app
YouTube imports
Pasting a YouTube URL contacts YouTube through yt-dlp and saves the selected source to your local app data.
YouTube connection and uploads
Connecting a channel uses your Google OAuth setup with YouTube upload and read-only permissions. Tokens are stored locally. Only videos and metadata you choose are uploaded. Optional outcome learning is off by default and, when enabled, reads public views, likes, and comments for uploaded videos.
Public game context
Game identification and title preparation can query Wikidata, Wikipedia, and Wikibooks. A query may be derived from an explicit game title, a source filename or folder, creator context, or an explicit game phrase in a transcript. Attributed text results can be cached locally; source media is not sent with those queries.
Model downloads and local AI
Whisper, YOLO, and optional SpeechBrain model files may be
downloaded from their model hosts when a required model is
missing. Inference on your media remains local. Optional
Ollama communication is limited to
127.0.0.1:11434 on your computer.
4. Controls and deletion
- You can clear the transcript cache from the app.
- Disconnecting a YouTube account removes its local token file.
- Share-safe feedback export removes paths, transcripts, private notes, identifiers, and timing details.
-
Reset and clear actions may first create local
.bakrecovery files. - Uninstalling the program does not automatically remove the app-data folder.
To remove all remaining local data, close ClipsAnyone? and delete
%LOCALAPPDATA%\ClipsAnyone, including any backup
files you no longer want.
5. Questions and changes
Privacy questions and product feedback can be raised in GitHub Discussions. This page may change as the beta adds, removes, or changes features. The updated date above will change when the description is revised.
The desktop app is open source. You can inspect its current implementation in the ClipsAnyone repository.