jiji262/douyin-downloader

Douyin Downloader V2.0

A practical Douyin downloader supporting videos, image-notes, collections, music, favorites collections, and profile batch downloads, with progress display, retries, SQLite deduplication, download integrity checks, and browser fallback support.

Desktop App (Douzy)

A desktop GUI built on the same backend, with dedicated workspaces for Douyin, TikTok, and YouTube. Paste a link to start, sync account content, follow every task, and manage downloaded works in a local archive.

• Three platforms: Douyin videos, galleries, profiles, and collections; TikTok videos, photos, and profiles; YouTube videos, Shorts, channels, and playlists
• Account content: sync Douyin following, favorites collections, collected series, and likes
• Visual workflow: multi-link queue, task status and retry controls, local download archive, filters, and quick re-download

Beta: The desktop app is currently in closed beta. To try it, download the build from the Releases page.

Douyin link download	TikTok download workspace	YouTube workbench
Paste a video, gallery, profile, or collection link and start in one click.	Download public videos, photo posts, and profiles without signing in.	Scan videos, Shorts, channels, and playlists, then configure video, MP3, or subtitle downloads.
Following management	Favorites and likes	Task Center
Sync creators, filter new works, add notes, and download directly from the list.	Browse collected videos, series, and liked works from the current Douyin account.	Track job results, retry failures, and open output folders.

Screenshots were captured from the current desktop main build. Demonstration data is used for privacy.

Feature Overview

⚠️ Douyin's anti-bot gate blocks the CLI from downloading likes / favorites / favorite collections (since 2026-08) and single videos / notes, collections and music (since 2026-09); profile posts can only rely on the browser fallback. See Current Limitations for the cause and what still works; use the Douzy desktop app for these downloads.

Supported

Current Limitations

• Douyin Argus gate: Douyin's edge ArgusSecurityPlugin answers every non-browser request to these endpoints with HTTP 403 Blocked by ArgusSecurityPlugin Uifid Not Found, with or without cookies and no matter how often you retry:

  • since 2026-08: aweme/favorite, collects/*, aweme/listcollection, mix/listcollection (likes / favorites / favorite collections)
  • since 2026-09-10: mix/aweme (collection items)
  • since 2026-09-14: aweme/detail (single video / note), aweme/post (profile posts), mix/detail, mix/list, music/detail, music/aweme, music/list

  The required x-secsdk-web-signature can only be produced by the SDK inside a real Douyin web page, which the CLI's direct API requests cannot carry, so single videos / notes, collections, music and likes / favorites cannot be downloaded in the CLI. Profile-post (post) API paging is rejected as well; with playwright installed and browser_fallback left on (headed by default), the browser fallback reads the page's own post-list requests and may still work, but it has not been tested against this gate. The Douzy desktop app sends these requests through its built-in login window and is not affected. Endpoints still reachable directly as of 2026-09-14: user profile, following list, comments, live rooms (webcast), hot board and search.

• Browser fallback is fully validated for post; like/mix/music currently relies on API pagination

• number.allmix / increase.allmix are retained as compatibility aliases and normalized to mix

• collect / collectmix currently work for the account represented by the logged-in cookies only

• collect / collectmix must be used alone and cannot be combined with post / like / mix / music

• increase currently applies to post / like / mix / music; favorites collection modes do not support incremental stop

• Live stream recording saves FLV natively; HLS sources only save the playlist (use ffmpeg for playable output)

• The webcast room endpoint is not verified against every live scenario — treat as experimental

Quick Start

1) Requirements

• Python 3.8+
• macOS / Linux / Windows

2) Install dependencies

pip install -r requirements.txt

For browser fallback and automatic cookie capture:

pip install playwright
python -m playwright install chromium

3) Copy config file

cp config.example.yml config.yml

4) Get cookies (recommended: automatic)

python -m tools.cookie_fetcher --config config.yml

After logging into Douyin, return to the terminal and press Enter. Cookies will be written to your config automatically.

5) Docker deployment (optional)

docker build -t douyin-downloader .
docker run -v $(pwd)/config.yml:/app/config.yml -v $(pwd)/Downloaded:/app/Downloaded douyin-downloader

Minimal Working Config

link:
  - https://www.douyin.com/user/MS4wLjABAAAAxxxx

path: ./Downloaded/
mode:
  - post

number:
  post: 0
  collect: 0
  collectmix: 0

thread: 5
retry_times: 3
proxy: ""
database: true
database_path: dy_downloader.db

progress:
  quiet_logs: true

cookies:
  msToken: ""
  ttwid: YOUR_TTWID
  odin_tt: YOUR_ODIN_TT
  passport_csrf_token: YOUR_CSRF_TOKEN
  sid_guard: ""

browser_fallback:
  enabled: true
  headless: false
  max_scrolls: 240
  idle_rounds: 8
  wait_timeout_seconds: 600

transcript:
  enabled: false
  model: gpt-4o-mini-transcribe
  output_dir: ""
  response_formats: ["txt", "json"]
  api_url: https://api.openai.com/v1/audio/transcriptions
  api_key_env: OPENAI_API_KEY
  api_key: ""

Usage

Run with a config file

python run.py -c config.yml

Append CLI arguments

python run.py -c config.yml \
  -u "https://www.douyin.com/video/7604129988555574538" \
  -t 8 \
  -p ./Downloaded

Arguments

Typical Scenarios

Download one video

link:
  - https://www.douyin.com/video/7604129988555574538

Download one image-note

link:
  - https://www.douyin.com/note/7341234567890123456

Download a collection

link:
  - https://www.douyin.com/collection/7341234567890123456

Download a music track

link:
  - https://www.douyin.com/music/7341234567890123456

Batch download a creator's posts

link:
  - https://www.douyin.com/user/MS4wLjABAAAAxxxx
mode:
  - post
number:
  post: 50

Batch download a creator's liked posts

link:
  - https://www.douyin.com/user/MS4wLjABAAAAxxxx
mode:
  - like
number:
  like: 0    # 0 means download all

Download multiple modes at once

link:
  - https://www.douyin.com/user/MS4wLjABAAAAxxxx
mode:
  - post
  - like
  - mix
  - music

Cross-mode deduplication: the same aweme_id won't be downloaded twice across different modes.

Download logged-in favorites collection items

link:
  - https://www.douyin.com/user/self?showTab=favorite_collection
mode:
  - collect
number:
  collect: 0

Download logged-in collected mixes

link:
  - https://www.douyin.com/user/self?showTab=favorite_collection
mode:
  - collectmix
number:
  collectmix: 0

Record a live stream (experimental)

link:
  - https://live.douyin.com/123456789   # or /follow/live/{room_id}
live:
  max_duration_seconds: 3600   # 0 = record until broadcaster ends
  chunk_size: 65536
  idle_timeout_seconds: 30

The recorder saves an FLV file under Downloaded/{author}/live/ plus a *_room.json metadata snapshot. If the broadcaster ends the stream, network goes idle, or you Ctrl+C, any already-recorded bytes are preserved (the .tmp file is promoted to the final file).

Collect comments per aweme

comments:
  enabled: true
  include_replies: false   # true will fetch each comment's second-level replies (extra API calls)
  max_comments: 500        # 0 = no cap
  page_size: 20

Generates a {date}_{title}_{aweme_id}_comments.json next to the media file.

Dump the hot search board

python run.py --hot-board 30 -p ./Downloaded
# Output: ./Downloaded/hot_board/20260424_221530.jsonl

Search by keyword

python run.py --search "猫咪" --search-max 100 -p ./Downloaded
# Output: ./Downloaded/search/猫咪_20260424_221530.jsonl

Run as REST API server

pip install fastapi uvicorn       # one-time optional dep
python run.py --serve --serve-port 8000

Endpoints:

Finished jobs are pruned by TTL (default 24h) and max-jobs (default 500) — in-flight jobs are never pruned. Configure via server.max_jobs / server.job_ttl_seconds.

Send a notification on completion

notifications:
  enabled: true
  on_success: true
  on_failure: true
  providers:
    - type: bark
      url: https://api.day.app/YOUR_DEVICE_KEY
      sound: bell
    - type: telegram
      bot_token: "123456:ABC..."
      chat_id: "987654321"
    - type: webhook                 # works with 企业微信/飞书/钉钉 bot URLs too
      url: https://qyapi.weixin.qq.com/cgi-bin/webhook/send?key=xxx
      extra_body:
        msgtype: text

All enabled providers are notified in parallel; a failing provider never blocks the download flow.

Incremental download (disk-based)

increase:
  post: true

With true, the downloader skips an item only when its non-empty primary media already exists under the current download directory. Deleting the media file makes the next run download it again; SQLite history does not affect this decision.

Set a mode to false to redownload and atomically replace existing files within the current number/date/media filters.

Full crawl (no item limit)

number:
  post: 0

Optional Feature: Video Transcription (transcript)

Current behavior applies to video items only (image-note items do not generate transcripts).

1) Enable in config

transcript:
  enabled: true
  model: gpt-4o-mini-transcribe
  output_dir: ""        # empty: same folder as video; non-empty: mirrored to target dir
  response_formats:
    - txt
    - json
  api_key_env: OPENAI_API_KEY
  api_key: ""           # can be set directly, or via environment variable

Recommended to provide key through environment variable:

export OPENAI_API_KEY="sk-xxxx"

2) Output files

When enabled, it generates:

• xxx.transcript.txt
• xxx.transcript.json

If database: true, job status is also recorded in SQLite table transcript_job (success/failed/skipped).

Testing

Recommended:

python3 -m pytest -q

Plain pytest is also supported now:

pytest -q

Key Config Fields

Output Structure

Default with folderstyle: true and database_path: dy_downloader.db:

workspace/
├── config.yml
├── dy_downloader.db          # default location when database: true
└── Downloaded/
    ├── download_manifest.jsonl
    ├── hot_board/                # when --hot-board is used
    │   └── 20260424_221530.jsonl
    ├── search/                   # when --search is used
    │   └── 猫咪_20260424_221530.jsonl
    └── AuthorName/
        ├── post/
        │   └── 2024-02-07_Title_aweme_id/
        │       ├── ...mp4
        │       ├── ..._cover.jpg
        │       ├── ..._music.mp3
        │       ├── ..._data.json
        │       ├── ..._avatar.jpg
        │       ├── ..._comments.json    # when comments.enabled
        │       ├── ...transcript.txt
        │       └── ...transcript.json
        ├── like/
        │   └── ...
        ├── mix/
        │   └── ...
        ├── music/
        │   └── ...
        ├── collect/
        │   └── ...
        ├── collectmix/
        │   └── ...
        └── live/                 # when recording live streams
            └── 2026-04-24_2215_LiveTitle_RoomId/
                ├── ...flv
                └── ..._room.json

Re-downloading Content

The program uses a database record + local file dual check to decide whether to skip already-downloaded content. To force re-download, you need to clean up accordingly:

Re-download a specific item

# Delete local files (folder name contains the aweme_id)
rm -rf Downloaded/AuthorName/post/*_<aweme_id>/

# Delete database record
sqlite3 dy_downloader.db "DELETE FROM aweme WHERE aweme_id = '<aweme_id>';"

Re-download all items from a specific author

rm -rf Downloaded/AuthorName/
sqlite3 dy_downloader.db "DELETE FROM aweme WHERE author_name = 'AuthorName';"

Full reset (re-download everything)

rm -rf Downloaded/
rm dy_downloader.db

Note: Deleting only the database but keeping files will NOT trigger re-download — the program scans local filenames for aweme_id to detect existing downloads. Deleting only files but keeping the database WILL trigger re-download (the program treats "in DB but missing locally" as needing retry).

FAQ

1) Why do I only get around 20 posts?

This is a common pagination risk-control behavior. Make sure:

• browser_fallback.enabled: true
• browser_fallback.headless: false
• complete verification manually in the browser popup, and do not close it too early

2) Why is the progress output noisy/repeated?

By default, progress.quiet_logs: true suppresses logs during progress stage.
Use --show-warnings or -v temporarily when debugging.

3) What if cookies are expired?

Run:

python -m tools.cookie_fetcher --config config.yml

4) Why are transcript files not generated?

Check in order:

• whether transcript.enabled is true
• whether downloaded items are videos (image-notes are not transcribed)
• whether OPENAI_API_KEY (or transcript.api_key) is valid
• whether response_formats includes txt or json

5) How to view download history?

sqlite3 dy_downloader.db "SELECT aweme_id, title, author_name, datetime(download_time, 'unixepoch', 'localtime') FROM aweme ORDER BY download_time DESC LIMIT 20;"

Community Group

点击链接加入群聊【QQ群】：https://qm.qq.com/q/9xoNt8Wzv4

Disclaimer

This project is for technical research, learning, and personal data management only. Please use it legally and responsibly:

• Do not use it to infringe others' privacy, copyright, or other legal rights
• Do not use it for any illegal purpose
• Users are solely responsible for all risks and liabilities arising from usage
• If platform policies or interfaces change and features break, this is a normal technical risk

By continuing to use this project, you acknowledge and accept the statements above.

License

This project is licensed under the MIT License. See LICENSE for details.

Friendly Links

• LINUX DO