- Home
- AI Video Search Tools
- Cinestar
Cinestar
What it is
Cinestar is a free, MIT-licensed application for searching images and videos with local AI. It can ingest folders or videos, transcribe audio with Whisper, embed frames with BGE, and create captions with Moondream. Natural-language searches return relevant moments with timecodes and thumbnails. Cinestar stores vectors in SQLite-Vec and uses Llama 3.2 to process and organize content. Its maker says AI processing runs locally, with no analytics, telemetry, usage tracking, or data uploads. The listed models are BGE embeddings, Moondream v2 for image and video captions, and Llama 3.2 (3B) for language tasks. Local AI services require Ollama, and the install guide lists model downloads totaling about 2–4 GB. The site describes plugins for face recognition, album creation, metadata extraction, transcoding, collaboration, LDAP, audit trails, backups, compliance reporting, and multi-user access, plus an open API for custom plugins. Linux and macOS are listed. The maker presents it for personal media managers, content creators, media organizations, and academic or market researchers.
For and against
For
- Searches images and videos with local AI
- Returns moments with timecodes and thumbnails
- Supports audio transcription and frame embeddings
- Open API supports custom plugins
- No analytics, telemetry, or data uploads, says maker
Against
- Local AI services require Ollama
- Model downloads total about 2–4 GB
- Listed platforms are Linux and macOS
The verdict
Cinestar combines local media indexing with natural-language search across images and videos. Its Ollama requirement and model downloads are worth considering before setup.
Who it suits
Cinestar may suit people managing personal media, creating content, or working in media organizations. Academic and market researchers are also among the maker's listed use cases.
Plans and prices
Cinestar (MIT open-source core)
Open-source core; local AI services require Ollama
Read 4 Oct 2026 · cinestar.sourceforge.io
On the record
- AI dependency
- The install guide requires Ollama and lists roughly 2–4 GB of model downloads.cinestar.sourceforge.io · read 20 Oct 2026
- API
- Cinestar’s site advertises an open API for creating custom plugins.cinestar.sourceforge.io · read 20 Oct 2026
- Audio processing
- Audio files are transcribed and analyzed for searchable content.github.com · read 20 Oct 2026
- Changelog
- Version 0.1.52, dated October 11, 2025, unified configuration files and fixed Whisper model downloading in packaged builds.cinestar.sourceforge.io · read 20 Oct 2026
- Developer setup
- Building from source requires Node.js 18 or newer, npm, and the commands npm install followed by npm run dev or npm run build.cinestar.sourceforge.io · read 20 Oct 2026
- Docker deployment
- The documentation supplies Docker Compose configurations for Ollama and Whisper services, including optional multi-GPU load balancing.cinestar.sourceforge.io · read 20 Oct 2026
- Documentation depth
- The install page includes health checks, Ollama API testing, service ports, memory reservations, and Nginx routing examples.cinestar.sourceforge.io · read 20 Oct 2026
- Downloads
- The v0.1.64 directory contains Linux ARM64, Linux x86_64, and macOS installers dated November 5, 2025.sourceforge.net · read 20 Oct 2026
- GitHub activity
- The public release repository shows 1 star, 1 fork, and zero listed issues or pull requests.github.com · read 20 Oct 2026
- Image recognition
- Images receive AI-generated captions that can be searched.github.com · read 20 Oct 2026
- Independent architecture detail
- The independent report says Cinestar combines 70% vector similarity from sqlite-vec with 30% SQLite FTS5 scoring.ikouchiha47.github.io · read 20 Oct 2026
- Independent performance detail
- That write-up reports approximately 40 seconds to make a 60-minute video searchable during its initial audio phase.ikouchiha47.github.io · read 20 Oct 2026
- Independent technical report
- An independent engineering write-up describes a five-phase pipeline that makes video segments searchable after local Whisper transcription and later enriches them with visual captions.ikouchiha47.github.io · read 20 Oct 2026
- Indexing
- Vectors are stored in SQLite-Vec and organized with Llama 3.2.cinestar.sourceforge.io · read 20 Oct 2026
- Installation
- The install guide instructs macOS users to copy Cinestar.app into Applications and clear Gatekeeper quarantine when necessary.cinestar.sourceforge.io · read 20 Oct 2026
- Integrations
- Listed plugin possibilities include Google Photos-style face recognition, social sharing, LDAP, audit trails, backups, compliance reporting, and multi-user access control.cinestar.sourceforge.io · read 20 Oct 2026
- Intended users
- The maker presents personal media managers, content creators, media organizations, and academic or market researchers as use cases.cinestar.sourceforge.io · read 4 Oct 2026
- Licence
- Cinestar is presented as open source under the MIT licence.sourceforge.net · read 20 Oct 2026
