HololiveToolkit/hololive Dreams

A toolkit for hololive Dreams.

Important Notice

As a courtesy to other fans, please refrain from spoiling unreleased content if any is found after decrypting the game data.

How to use

1. Install the requirements.

   python -m pip install -r requirements.txt

   Video conversion also requires ffmpeg on PATH. Audio conversion uses the cridecoder Python package installed from requirements.txt.

2. Copy octocacheevai into the cache folder.

   The default Steam location is:

   *\steamapps\common\hololiveDreams\octo\pdb\5\400001\octocacheevai
   

   The resulting toolkit layout should be:

   hololive-toolkit\
   ├─ cache\
   │  └─ octocacheevai
   ├─ main.py
   └─ requirements.txt
   

3. Run main.py.

   By default, this downloads and extracts every category for Japanese and shared data:

   python main.py

   Use --language to select another language. For example, use kor for Korean:

   python main.py --language kor

   --cache and --output are optional. By default, the toolkit reads cache/octocacheevai, caches Unity bundles in cache/bundles, caches converted media sources in cache/resources, and writes only final extracted files to cache/extract.

Languages and categories

Supported languages:

jpn, eng, kor, chs, cht, ind, all

• jpn: Japanese (default)
• eng: English
• kor: Korean
• chs: Simplified Chinese
• cht: Traditional Chinese
• ind: Indonesian
• all: All languages

Supported categories:

img, adv, live2d, model, motion, effect, audio, video, chart, all

Omitting --categories is the same as --categories all. Specify a comma-separated list only when limiting the download.

audio includes voice, BGM, SE, and other ACB/AWB resources. External AWB files and AWBs embedded in ACB files are decoded to PCM16 WAV automatically. Non-Japanese language selections include shared base assets together with the matching _lang-* assets.

Examples

Inspect the local OCTO database:

python main.py --inspect

Write the decoded manifest as JSON:

python main.py --inspect --json "cache\OctoManifest.json"

Extract a small test selection:

python main.py --categories img,adv --limit 10 --workers 4

Download all categories for all languages:

python main.py --language all

Download and decrypt without extracting Unity objects or converting media:

python main.py --no-extract

Download and extract voice, BGM, and sound effects to WAV:

python main.py --categories audio

Download, decrypt, and convert CRI USM videos to MP4:

python main.py --categories video

The source .usm is removed only after the generated MP4 passes a decode validation. Hololive Dreams' videos do not require a CRI Movie key; --movie-key is available for compatible encrypted data.

Existing downloads are skipped automatically and can be resumed with the same command. Use --overwrite to download them again.

Output

Only final extracted files are stored in category folders. Model textures use a dedicated textures subdirectory:

cache\extract\
├─ img\*.png
├─ adv\*.json
├─ live2d\*
├─ model\*
│  └─ textures\*.png
├─ motion\*
├─ effect\*
├─ voice\wav\*.wav
├─ bgm\wav\*.wav
├─ se\wav\*.wav
├─ audio\wav\*.wav
├─ video\*.mp4
└─ chart\*

Downloaded Unity bundles are retained once under the shared cache/bundles directory, so repeated downloads and extraction reuse the same files. Use --bundle-cache to choose another location. Only embedded model Texture2D and Sprite objects are exported under cache/extract/model/textures.

Original ACB/AWB files are retained under cache/resources for reuse. Use --resource-cache to choose another location. Single-entry banks produce wav/<bank>.wav; multi-entry banks use wav/<bank>/<bank>_<id>.wav.

Special Thanks

• vilebbit/HoshimiToolkit
• vilebbit/vision-meta