whatsthis

Pure Go library that guesses video metadata (title, year, codec, resolution, release group, etc.) from a filename. Inspired by the Python guessit project.

Installation

go get github.com/odwrtw/whatsthis

Usage

info := whatsthis.Video("Big.Buck.Bunny.2008.1080p.BluRay.x264-YIFY.mkv")
fmt.Println(info.Type)         // movie
fmt.Println(info.Title)        // Big Buck Bunny
fmt.Println(info.Year)         // 2008
fmt.Println(info.ScreenSize)   // 1080p
fmt.Println(info.VideoCodec)   // H.264
fmt.Println(info.ReleaseGroup) // YIFY
fmt.Println(info.Container)    // mkv

Use File as the general entry point for extensionless release names and video files. It recognizes whole-season names before falling back to Video:

info := whatsthis.File("Velvet Meridian 2026 Season 1 Complete 1080p WEB x264 [theta_group]")
fmt.Println(info.Type)    // show_season
fmt.Println(info.Title)   // Velvet Meridian
fmt.Println(info.Season)  // 1
fmt.Println(info.Episode) // 0

Explicit multi-season ranges are returned as a sorted, deduplicated Seasons list:

info := whatsthis.File("Example.Show.S02-S04.1080p.x265-GROUP")
fmt.Println(info.Type)    // show
fmt.Println(info.Seasons) // [2 3 4]
fmt.Println(info.Season)  // 0

Seasons is nil for movies, episodes, and single-season releases, and omitted from JSON when nil. Overlapping ranges are merged; gaps between separate ranges are preserved. Descending ranges are not expanded. As with single-season packs, collection detection applies to extensionless release names, not video filenames.

Languages

Languages is a slice of language.Tag from golang.org/x/text/language. It combines explicit audio and subtitle language tags (for example, FRENCH, ENG, uppercase FR, VOSTFR, and SweSub) in encounter order without duplicates. CHS and CHT preserve Chinese script information as zh-Hans and zh-Hant.

info := whatsthis.Video("Example.2024.FRE.ENG.CHS.1080p.x264-GROUP.mkv")
fmt.Println(info.Languages)             // [fr en zh-Hans]
fmt.Println(info.Languages[0].String()) // fr

Compare tags with the predefined values such as language.French. For human-readable names, use language/display in the consuming application; English names are presentation, not identifiers.

API change: Languages is now []language.Tag, not []string. JSON and the WASM API serialize the tags as BCP 47 strings (for example, "languages":["fr","en","zh-Hans"]), not English names. Missing languages remain nil / JSON null.

Multilingual indicates multiple distinct tags (including script variants) or a generic MULTi, Multiple Subtitle, DUAL, or DL marker; WEB-DL alone does not set it. These fields do not establish which audio tracks are present, and unspecified languages are not inferred. Release-name recognition remains limited to the supported vocabulary and contextual rules, rather than interpreting every valid ISO code as language metadata.

See parsing policies and limitations for supported conventions and ambiguity handling.

WebAssembly demo

The static web interface runs whatsthis.File entirely in the browser; names are not uploaded anywhere.

Try it at odwrtw.github.io/whatsthis.

Build it with the repository's Go version:

./scripts/build-web.sh

Then serve dist with any static HTTP server and open index.html. The generated directory contains the site assets, the WebAssembly binary, and the matching wasm_exec.js from the local Go toolchain.

Pushes to master deploy the site through GitHub Actions. To enable deployment, select GitHub Actions as the source under Settings → Pages in the repository.