What is scraping, and how does Fushi's scraping work?

What scraping is ​

You have a file on disk called

[VCB-Studio] Sousou no Frieren [01][Ma10p_1080p][x265_flac].mkv

To a computer that's just a string. Scraping is: first recognise from the filename that this is Frieren: Beyond Journey's End, episode 1, then fetch the show's poster, synopsis, rating, season and episode lists and airing dates from an online database and store them locally. Afterwards your video library turns from a pile of filenames into a wall of covers, and opening a show gives you every episode's title and thumbnail.

The term comes from media-library software like Jellyfin / Plex / Kodi (a "scraper" scrapes information off web pages); the databases are sites like TMDB, MAL, AniDB and Bangumi.

What Fushi scrapes ​

Video is where scraping matters most; the rest of this article is mainly about it.

Video: automatic by default ​

Importing a local folder offers two ways of organising it:

• By series (default): recognise series and episodes from filenames, then match the show's data — this scrapes.
• By folder: one collection per top-level subfolder, organising only. For folders sorted by difficulty, course or topic.

With "By series", Fushi starts scraping as soon as the scan finishes (the source's "Scrape after scan" and the global "Auto-scrape item details" are both on by default); videos from anime downloads are scraped before they enter the library too (download → organise → subtitles → scrape → library). One more step follows: Auto-fetch subtitles after scraping grabs a subtitle from your configured online sources for videos that still have none.

Scraping works by confidence: only a single exact hit for the parsed title is applied automatically; several same-named candidates, or no hit on either side, and it doesn't guess — the item goes to Pending confirmation for you to decide. The video library shows a banner "N shows haven't had their identity confirmed" with a "Go confirm" button.

Pending confirmation: pick the show yourself ​

The Background tasks panel on the video source page has three tabs: Current tasks / Pending confirmation / Recent tasks. Under "Pending confirmation", open a show:

• By title: results come with covers and years; selecting one binds it;
• By ID: choose MAL, TMDB film or TMDB TV, enter the ID, or simply paste the work's official page URL.

A wrongly scraped show can be fixed too: collection menu → Re-scrape details and cover → pick the right entry. A manual binding is protected — later automatic runs won't change it; covers you set yourself (a local image you chose, a poster that came with the folder) are never overwritten either.

If a single episode lands on the wrong season / episode, use Pin to season/episode… in that episode's menu; the pin is kept on every later scrape. Several seasons mixed into one collection? The collection menu has Split by season.

Naming files so they're recognised ​

Fushi's filename parser follows the Jellyfin / anitomy line: it strips fansub tags, quality and codec, and understands S01E01, 第2季, 2nd Season, Season 2. A reliable layout:

Anime/
  Sousou no Frieren/
    [VCB-Studio] Sousou no Frieren [01][Ma10p_1080p].mkv
    [VCB-Studio] Sousou no Frieren [02][Ma10p_1080p].mkv
  Hibike! Euphonium/
    Season 2/
      Hibike! Euphonium - 01.mkv

A few rules of thumb:

• one folder per show; the season number can live on the subfolder alone (Season 2 / S02 / 第2季);
• zero-pad episode numbers (01, not 1) — whether Show 2.mkv is season 2 or episode 2 is a coin toss, though for a batch like Show 1 … Show 12 in one folder Fushi works out the episode numbers by comparing the siblings;
• use the original or official English title rather than a nickname you made up;
• if a filename won't parse and you don't want to rename, use Identifier rules (below) or turn on AniDB file recognition.

Advanced settings ​

Primary source, identifier words and metadata language are under Settings → Video → Library; the AniDB account and TMDB API key are under Settings → Online services → Metadata scraping; and each source has its own Source scrape settings:

• Primary metadata source: AniDB (hash identity + anime XML) by default, with seasons and episodes organised the AniDB way and TMDB always supplementing; MAL (via Jikan) or TMDB can be chosen instead. When the primary has no single exact hit or is unavailable, another source fills in automatically. Switching the primary source doesn't re-identify works that are already recognised. The metadata language (synopsis, TMDB poster language) can be set per source.
• Identifier rules: rewrite filenames before recognition, one rule per line — blocking is a plain regex; replacement A => B; episode offset before <> after >> EP+1 (for example shifting a second season that numbers from 13 back to 1).
• Recognise files via AniDB ED2K: identify show and episode from the file's hash on AniDB — works however mangled the filename is. Needs your own AniDB account; AniDB's UDP login is unencrypted, so only enable it on a trusted network.
• TMDB API key: one is built in; enter your own only if scraping breaks or you want your own quota.
• NFO and image write policy: scraped data can be written as .nfo files and poster.jpg next to the videos — Jellyfin / Kodi read the same format. Default "Write only when missing"; third-party files and ones you edited are protected and never overwritten.
• Locked fields: lock title, cover and other fields in the work's details so a re-scrape keeps your values.
• Clear all scrape records: wipes details, bindings and the NFO / covers Fushi generated; video files, watch progress and subtitles stay.

Books, manga, games ​

• Books / manga: Match cover online in the edit dialog searches Bangumi by title / author; pick one to use it.
• Games: Scrape metadata from the card menu searches VNDB / Bangumi by the current name as soon as it opens; candidates show covers, and "Use" writes cover and synopsis together.
• Scrape all in the library menu runs in bulk: videos are scored on title, year and more; books and games require a unique, exact title match to apply automatically; the rest go to pending confirmation.

With interconnect ​

When a phone watches remote videos from a PC host, the host's scraped data is authoritative and syncs to the phone automatically; for a show that wasn't scraped yet, the phone can choose Scrape on host or Scrape locally and write back to host. See Fushi interconnect.

Common situations ​

• Wrong show recognised: pick it manually under pending confirmation / re-scrape; for a filename that keeps failing, add an identifier rule.
• One show split into several: usually inconsistent season / episode notation in the filenames (S02E01 here, - 13 there) — unify them or use the episode offset rule; pin individual episodes with "Pin to season/episode…".
• Poster in the wrong language: change the source's "Metadata language".
• Search fails / rate limited: AniDB and MAL (via Jikan) are both rate limited, and AniDB temporarily bans clients that ask too often, so bulk scraping being a bit slow is normal; if it can't connect for now, try again a bit later.
• Scraping vs. "Bangumi sync": scraping pulls data in; Bangumi sync pushes your watching / reading progress to your Bangumi account.