What we do
ShanHaiYin supports image, text, audio, and video analysis. Images are checked for available provenance credentials before generation, synthesis, and editing indicators are evaluated. Text can be pasted or uploaded as a plain-text file. Audio and video are analyzed as complete media files so the model can evaluate continuous sound or frame evidence. Each completed analysis provides a primary finding, an estimated AI likelihood, supporting signals, a request reference, and a SHA-256 content fingerprint.
Our principles
- Verify direct, machine-readable provenance before relying on probabilistic model analysis.
- Explain the finding and key evidence instead of presenting an isolated score.
- Record request references and SHA-256 fingerprints so reports can be matched to source content.
- State the limits of each content type and avoid treating a video screenshot as a substitute for full video analysis.
Service boundaries
AI content detection is affected by file quality and processing history. Screenshots, heavy compression, cropping, transcoding, and substantial editing may weaken original signals. Important content should be checked using the original file whenever possible and considered together with creation records, publication history, and source information.