Privacy Policy for Captio
Effective Date: October 22, 2026
Applicable Product: Captio Video Subtitle Editor (Android)
Privacy-First Guarantee: Captio processes your video recordings, audio tracks, speech-to-text transcriptions, and subtitle texts entirely on your device. No media files, voice samples, or subtitle transcripts are ever uploaded to cloud servers or remote speech recognition services.
1. Developer Identification & Contact
Captio is developed and operated by Vladimir Baev (“Developer”, “we”, “us”, or “our”).
If you have questions, feedback, or data requests regarding this Privacy Policy, please contact:
2. Local On-Device Media & Subtitle Processing
Captio provides video editing, automated speech recognition, caption styling, and subtitle synchronization. All core computational workloads execute strictly on the client hardware:
- Video and Audio Files: When you import or record a video, audio decoding, waveform extraction, video preview, and MP4 rendering occur strictly inside local device memory using Android platform codecs and local native libraries (such as Mp4Composer). No media data is streamed or transmitted over the internet.
- Automated Speech Recognition (Whisper): Automated speech-to-text transcription uses quantized OpenAI Whisper neural models running locally on your device’s CPU via whisper.cpp. Voice audio is sliced into temporary local WAV/PCM files stored in private app cache, transcribed on-device, and immediately deleted upon completion.
- Word-Level Alignment (Whistle): When optional Russian phoneme word-alignment is enabled, local neural inference is executed entirely on-device via Microsoft ONNX Runtime CPU. No audio or alignment data leaves the device.
- Subtitle Translation (Google ML Kit): Subtitle translation between languages is executed locally on your device using Google ML Kit On-Device Translation. Your subtitle text remains on-device and is never sent to remote Google Cloud Translation servers.
3. Network Connections & External Downloads
Captio is fully functional offline after required assets are downloaded. Network connections are used for downloading requested resources, retrieving app configuration, and sending analytics and diagnostic telemetry. These connections use HTTPS:
- Speech Model Downloads: Quantized Whisper speech models are downloaded from public model repositories (Hugging Face / GitHub) only when you initiate subtitle generation or download a model in Settings.
- On-Device Translation Language Packages: Google ML Kit downloads on-device translation language packages (~30 MB per language pair) directly from Google servers over HTTPS upon first user request.
- Downloadable Fonts: When you select an optional subtitle font from the in-app catalog (such as Ubuntu, PT Sans, Montserrat Alternates, Playfair Display SC, or Fira Mono), the font is downloaded directly from Google Fonts over HTTPS and stored locally in the app’s private files.