hbo — HBO Max (Adria) → Serbian-learning catalogue

A self-contained web UI for browsing the HBO Max (Serbia / Adria region) catalogue with faceted search and per-title audio/subtitle-language filtering — built as a tool for learning Serbian by finding what's watchable in Serbian, Croatian, Bosnian (and Russian) audio or subtitles.

---

Status: one-off personal tool, complete & live · Stack: vanilla HTML/CSS/JS (no framework, no dependencies, no backend) + a few Python/JS harvest scripts · Live at: https://den.tanaev.su/hbo/ · Data: embedded as a static site/catalog.json (~1.8 MB).

---

What it does

HBO Max in the Adria region carries a lot of Balkan content and near-universal SR/HR subtitles, but the player gives you no way to ask "which titles actually have Serbian audio?" or "what's dubbed vs. native?". This project answers that question offline.

It's a single-page catalogue of ~2,405 titles (1,386 movies + 1,019 series, harvest of 2026-08-29) pulled from HBO Max's content API for a logged-in Serbia account. For every title it stores metadata only — title, description, cover-image URL, year, seasons/episodes, genres, age rating — plus the complete list of audio and subtitle languages. The web UI then lets you filter by those languages, so a Serbian learner can zero in on native ex-Yugoslav productions, Serbian/Croatian dubs, or anything with SR/HR subtitles.

The learning-Serbian angle drives the UI: Serbian, Croatian, Bosnian and Montenegrin (the mutually intelligible "BCS" family), plus Russian, are pinned to the top of the language facets and ★-highlighted everywhere. Cards flag priority-language availability with AUD / SUB badges.

Scope: metadata + cover-image URLs only — enough to browse what is available. It does not download video (DRM-protected). Covers load live from HBO's public CDN, so the UI needs internet to show thumbnails.

Screenshot / example

No image is committed. To see it, open https://den.tanaev.su/hbo/ or serve site/ locally (below). The UI is a dark, poster-grid catalogue: a search bar up top, a scrollable left sidebar of facets (Тип / Год / Озвучка-аудио / Субтитры / Жанр), a poster grid with live result counts, and a detail modal per title listing every audio & subtitle language with a "Открыть на HBO Max" link. The UI labels are in Russian (the owner's working language); the catalogue data is English titles.

An example row from site/catalog.json:

{
  "id": "020b4de1-…", "type": "movie", "title": "33 Photos from the Ghetto",
  "year": 2026, "rating": "12", "genres": ["documentary"],
  "cover": "https://images.cdn.prd.api.discomax.com/…jpeg",
  "audio": ["Polish - Original", "English"],
  "subs": ["English - CC", "Croatian", "Serbian (Latin)", "Russian", …],
  "seasons": 1, "episodes": 7, "kids": false, "pop": 0.0,
  "url": "https://play.hbomax.com/movie/020b4de1-…"
}

Architecture

Two phases: a manual harvest → build pipeline that produces site/catalog.json, and a static site that reads that JSON entirely in the browser. There is no server-side logic and no runtime API call — once catalog.json exists, the site is just three static files.

  HBO Max content API                       (harvest, in a logged-in Chrome tab,
  default.any-emea.prd.api.hbomax.com         driven by Claude-in-Chrome javascript_tool)
        │  cookie auth (credentials:'include') + the app's own client headers
        ▼
  scripts/00-api-headers.js         ──► window.__H  (the x-disco-*/x-device-info headers the SPA sends)
  scripts/01-enumerate-catalog.js   ──► window.localStorage['hbocat_v1']  (resumable state)
  scripts/02-harvest-languages.js   ──►   { titles: { id: {name, audio[], subs[], …} } }
  scripts/01c-enrich-ratings.js     ──►
        │
        ▼  scripts/03-export-base64.js  →  COPY button  →  (real click)  →  macOS clipboard
        │                                                    │  pbpaste | base64 -D
        ▼                                                    ▼
  scripts/06-add-popularity.py   ── adds `pop` from TMDb keyless export ──►  site/catalog.json
  scripts/04-build-from-json.py  ── derives CSVs ──►  data/languages-matrix.csv, hbo-catalog-full.csv
        │
        ▼
  site/index.html  ── fetch('./catalog.json') ──►  faceted UI (all filtering client-side, in-browser)
        │
        ▼  scripts/05-deploy.sh  (rsync over ssh)
  nginx static host  ──►  https://den.tanaev.su/hbo/

External systems it touches: HBO Max / WBD "Beam" content API (read-only, cookie auth, harvest only); TMDb's public keyless popularity export (files.tmdb.org, no API key); HBO's public image CDN (images.cdn.prd.api.discomax.com, at runtime for thumbnails); the nginx host at den.tanaev.su (deploy target).

See HANDOFF.md for the full reverse-engineered HBO Max API playbook.

Getting started (run locally)

No install, no dependencies. Just serve the site/ directory over HTTP (a plain file:// open won't work because the page does fetch('./catalog.json')).

git clone https://github.com/deetan/hbo.git
cd hbo
python3 -m http.server 8766 --directory site
# then open http://localhost:8766/

Thumbnails load from HBO's public CDN, so you need internet to see cover images; the catalogue itself (titles, languages, filtering) works fully offline once the page has loaded catalog.json.

Prerequisites: any static file server (Python 3's http.server is enough). Regenerating the catalogue (below) additionally needs Chrome logged into HBO Max in the Serbia region + the Claude-in-Chrome tooling, and Python 3 for the two build scripts.

Regenerating the catalogue

HBO's catalogue changes over time, so site/catalog.json goes stale. To rebuild it you re-harvest from a live, logged-in HBO Max browser session. The full step-by-step (with API details and gotchas) is in HANDOFF.md → "To REFRESH the catalogue later"; the short version:

1. Chrome open and logged into HBO Max, region Serbia, on a play.hbomax.com tab; connect Claude-in-Chrome to it.
2. Paste scripts/00-api-headers.js first: the API rejects plain cookie-auth calls with HTTP 400 unless they carry the x-disco-client / x-disco-params / x-device-info headers the web app sends, so this script borrows them from one of the app's own requests into window.__H.
3. Wipe the previous harvest with localStorage.removeItem('hbocat_v1'). Skipping this makes the run a top-up rather than a refresh — titles HBO has dropped would stay in the catalogue forever.
4. Paste, in order, into the page via javascript_tool: scripts/01-enumerate-catalog.js → scripts/02-harvest-languages.js (poll with scripts/02b-progress.js until finished) → scripts/01c-enrich-ratings.js. State accumulates in localStorage['hbocat_v1'] and the sweep is resumable.
5. scripts/03-export-base64.js injects a COPY button and returns its {x,y}. Do a real click on it (a trusted gesture is required to grant clipboard access), aiming with a fresh screenshot, then on the host: pbpaste | base64 -D > site/catalog.json. Verify it parses and count ≈ 2,400.
6. python3 scripts/06-add-popularity.py — adds the pop score (used by the default "most popular" sort) by matching titles against TMDb's keyless export.
7. python3 scripts/04-build-from-json.py — regenerates the two CSVs in data/ from the JSON.
8. Deploy (below).

The scripts are pasted into the browser rather than run with Node — see HANDOFF.md §6 for why (the site blocks programmatic downloads; the clipboard is the working data-out bridge).

Project layout

hbo/
├── README.md                       ← this file
├── CLAUDE.md                       ← agent-facing working notes (read this + HANDOFF next)
├── HANDOFF.md                      ← full context: HBO API playbook, gotchas, prior-pass results
├── .gitignore                      ← ignores .DS_Store
├── site/                           ← the deployable static site
│   ├── index.html                  ← the entire UI (HTML+CSS+JS, self-contained, ~25 KB)
│   └── catalog.json                ← embedded catalogue data (~1.8 MB, ~2,405 shown titles)
├── data/                           ← derived / source data (not needed by the site at runtime)
│   ├── hbo-catalog-full.csv        ← flat CSV of every title (built by script 04)
│   ├── languages-matrix.csv        ← language → #titles-with-audio / #titles-with-subs (script 04)
│   ├── hbo-max-sr-hr-AUDIO.md      ← earlier pass: the SR/HR-audio titles categorised (native vs dub)
│   └── hbo-max-sr-hr-catalog.csv   ← earlier audio-focused CSV (superseded by hbo-catalog-full.csv)
├── scripts/                        ← the harvest → build → deploy pipeline
│   ├── 00-api-headers.js           ← [browser] capture the client headers the API now demands
│   ├── 01-enumerate-catalog.js     ← [browser] enumerate all titles + metadata + cover URLs
│   ├── 01c-enrich-ratings.js       ← [browser] add age/maturity ratings
│   ├── 02-harvest-languages.js     ← [browser] per-title audio/subtitle lists + seasons/episodes
│   ├── 02b-progress.js             ← [browser] poll harvest progress
│   ├── 03-export-base64.js         ← [browser] build catalog.json + COPY-to-clipboard button
│   ├── 04-build-from-json.py       ← [host] derive the two CSVs from catalog.json
│   ├── 06-add-popularity.py        ← [host] add `pop` score from TMDb keyless export
│   └── 05-deploy.sh                ← [host] rsync site/ to the nginx host
└── .claude/
    └── launch.json                 ← local "hbo-catalog" server (python http.server on :8766)

Deploy

The site is static files under an existing nginx web root, so shipping is a plain rsync — no nginx config change:

bash scripts/05-deploy.sh

It rsyncs site/index.html + site/catalog.json to the den.tanaev.su host (an nginx static host, reached via the reports ssh alias) into …/public/hbo/, fixes perms, then curls /hbo/ and /hbo/catalog.json to confirm both return 200. The page sets noindex,nofollow (it's a personal tool on a public domain). Requires an ssh reports alias in your ~/.ssh/config with access to that host.

License / ownership

Personal project — private, © dt. HBO Max catalogue metadata and cover images are © Warner Bros. Discovery and are used here for personal, non-commercial reference only.