Every second leaves a mark. Shinhera extracts the unique visual DNA of any stream, providing absolute proof of content in milliseconds.
Developed by Karanfil Labs, Shinhera transforms passive screen viewing into interactive intelligence. Capture any movie or series scene via camera or screen radar to identify productions instantly and engage in real-time voice conversations with our AI film critic and envoy, Tiffany, across 24 languages.
Identifies movies and TV series directly from live camera frames or screen captures, delivering instant metadata and voice-guided cinematic analysis.
Natural, two-way voice dialogue with Tiffany featuring contextual long-term memory, personalized film discovery, and expressive speech synthesis in 24 languages.
Microphone audio and visual radar captures are processed in real time and permanently deleted immediately after analysis, with zero public AI model training.
Shinhera is engineered from the ground up by Karanfil Labs combining native mobile IDEs, on-premise speech recognition, real-time visual scene recognition, and dedicated Linux infrastructure.
Android app developed in Android Studio with Kotlin, Media3 ExoPlayer, AudioRecord, and OkHttp3/Retrofit2. iOS app built in Xcode with SwiftUI, AVQueuePlayer, and AVAudioEngine. Web portal powered by HTML5, Tailwind CSS, and JavaScript.
Identifies movie scenes and posters in real time through optical visual matching, enriched with live TMDB cinema metadata, cast profiles, and multilingual discovery.
On-premise Faster-Whisper (turbo int8 + VAD) transcribes speech directly in RAM. Google Gemini Flash (GenAI SDK) powers reasoning and memory summarization, while ElevenLabs (eleven_turbo_v2_5) synthesizes ultra-low-latency voice.
Hosted on Hetzner Ubuntu Linux servers with Python Flask, multi-worker Gunicorn, and Nginx. PostgreSQL manages ACID user archives and Quantum Credits, secured by Firebase Auth, Google reCAPTCHA Enterprise, and RevenueCat.