🇺🇦 Інструкція українською. This voice pack lets your robot vacuum speak using the voice of GLaDOS — the artificial intelligence from Portal 2, known for her sarcasm, superiority complex, and dark humor.
The project was made possible with financial support from the Ukrainian community.
The phrases have been rewritten with GLaDOS's signature dark humor and toxic personality — without losing the original meaning.
| Brand | Package Format | Supported Models |
|---|---|---|
| Roborock | .pkg | First- and second-generation robots: v1, M1S, S4, S5, S5e, S6, S6 MaxV, S6 Pure, T6 |
| Dreame | tar.gz | All dreame.vacuum models that support voice package installation via the specification |
| Xiaomi | tar.gz | All xiaomi.vacuum models that support voice package installation via the specification |
| Midea | .zip | Midea M7 Pro |
| Xiaomi Ijai | .zip | ijai.vacuum v2, ijai.vacuum v3, ijai.vacuum.v18, ijai.vacuum.v19 |
You can install the GLaDOS voice pack in several ways. A detailed guide is available here.
| url | https://github.com/sverdlyuk/glados_ukr/raw/refs/heads/main/uk_glados_voice_pack.gz |
|---|---|
| md5 | 3545e91c0626beccbd284469f6283a77 |
| size | 9620968 |
| Download sample voice |
- Install the Dreame vacuum integration.
-
In Home Assistant, go to
"Developer Tools"->"Services"and switch toYAML mode -
Paste the following code:
service: dreame_vacuum.vacuum_install_voice_pack data: lang_id: UK url: >- https://github.com/sverdlyuk/glados_ukr/raw/refs/heads/main/uk_glados_voice_pack.gz md5: 3545e91c0626beccbd284469f6283a77 size: 9620968 target: entity_id: vacuum.mops
-
Replace the
entity_idvacuum.mopswith theentity_idof your vacuum cleaner. -
Click the Run Action button.
- Install the integration Xiaomi Miot Auto
In Home Assistant, go to Developer Tools → Services and switch to YAML mode.
Paste the following code:
service: xiaomi_miot.set_miot_property
data:
entity_id: vacuum.dreame_p2041o_796c_robot_cleaner
siid: 7 # siid for the audio service
piid: 4 # piid for setting the voice pack
value: '{"id":"UK","url":"https://github.com/sverdlyuk/glados_ukr/raw/refs/heads/main/uk_glados_voice_pack.gz","md5":"3545e91c0626beccbd284469f6283a77","size":9620968}'Replace the entity_id vacuum.dreame_p2041o_796c_robot_cleaner with the entity_id of your robot vacuum.
Install python-miio and run the following command:
python -m miio.cli genericmiot --ip 192.168.50.157 --token 614a498f6c72506d6e3066764f73696a raw_command set_properties "[{'did': '8023334994', 'siid': 7, 'piid': 4, 'value' : '{\"id\":\"UK\",\"url\":\"https://github.com/sverdlyuk/glados_ukr/raw/refs/heads/main/uk_glados_voice_pack.gz\",\"md5\":\"3545e91c0626beccbd284469f6283a77\",\"size\":9620968}'}]"miiocli device --ip 192.168.50.157 --token 614a498f6c72506d6e3066764f73696a raw_command set_properties "[{'did': '8023334994', 'siid': 7, 'piid': 4, 'value' : '{\"id\":\"UK\",\"url\":\"https://github.com/sverdlyuk/glados_ukr/raw/refs/heads/main/uk_glados_voice_pack.gz\",\"md5\":\"3545e91c0626beccbd284469f6283a77\",\"size\":9620968}'}]"Replace the IP and token with your own.
- Open the Valetudo web interface by entering the IP address of your robot vacuum into your web browser.
- In Valetudo, go to "Robot Settings" -> "Misc Settings."
- Paste the following parameters into the "Voice packs" section:
- URL: https://github.com/sverdlyuk/glados_ukr/raw/refs/heads/main/uk_glados_voice_pack.gz - Language Code: UK - Hash: 3545e91c0626beccbd284469f6283a77 - File size: 9620968 - Save and click "Set Voice Pack".
💡 Lifehack:
If you take the link
https://awsde0.fds.api.xiaomi.com/dreame-product/dreame.vacuum.p2008/voices/soundpackage.json
and replace the model part (dreame.vacuum.p2008) with your own, you can see all the voice packages available for your vacuum cleaner.
This link works only fordreame.vacuummodels. You can find your model in the specifications. This way, you can easily view and download all officially available voice packages.
-
Prepare
.oggfiles, 16,000 Hz, Stereo/Mono according to the phrase table. You can create only 5-10 of the most popular phrases. Other phrases will be taken from the donor pack. -
Download any existing voice pack: GLaDOS_ukr, uk_female_pensive, dreame_female_uk, Les Poderevyansky. A voice pack is a tar.gz archive. For example, the GLaDOS_ukr voice pack has a .gz extension. To open the archive, rename the extension from
.tartotar.gz. Similarly, you can add the extension to a pack that has no extension. Open the archive (on MacOS I used BetterZip) and replace the archive files with your own. Note: the archive must contain only files, not a folder with files. -
Upload the voice pack somewhere online to get a link to the file. For example, a public repository on GitHub, GitLab, FTP server, etc. Find out the file size in bytes and the md5 checksum of the file. If you don’t know how to do this, you can use online services.
Demo video. This voice pack is intended to work on Roborock robot vacuums of the 1st and 2nd generation (v1, M1S, S4, S5, S5e, S6, S6_MaxV, S6_Pure, T6). I tested it on v1 (rockrobo.vacuum.v1, same as Gen 1 Xiaomi Mi SDJQRO2RR). Third-generation vacuums have voice packs signed with the Roborock certificate. Because of this, installing custom voice packs is not possible.
To install the voice pack:
- Download roborock_glados.pkg to your computer;
- Install python-miio;
- In the command line, in the folder with roborock_glados.pkg, run the command
mirobo --ip 192.168.0.124 --token 50597y49p1894f6d6e67477349747408 install-sound roborock_glados.pkgwhere 192.168.0.124 should be replaced with the IP address of your robot vacuum, and 50597y49p1894f6d6e67477349747408 with your robot vacuum's token. You can find the token using the command python -m miio.cli cloud list (Windows), miiocli cloud list (Linux, MacOS), or any other convenient method like this one.
To create your own voice pack, you need to:
- Create or download files in .wav format, bitrate 1411.2 KB/s, sample rate 44.1 or 16.0 kHz;
- Download the encryption utility ccrypt;
- In the command line, in the folder with the files, run the command
tar zc *.wav | ccrypt -e -K "r0ckrobo#23456" > roborock_glados.pkg
where roborock_glados should be replaced with any name you want for your voice pack.
6. Done! You can upload the created .pkg file to your robot vacuum.
- Download SDK Platform Tools to your PC and extract them to any convenient folder.
- Connect the robot to your PC. The USB port is located under the top cover beneath the Reset button, hidden behind a plug.
- Open the command line in the extracted Platform Tools folder (right-click -> Open in Terminal) and run the command
adb devicesto make sure your device is recognized. If you see your device ID, everything is fine; if not, your cable may not support data transfer (I tried several cables I had and only one worked). - Then download the voice pack and extract it to any folder convenient for you.
- Create a text file in the Platform Tools folder with the following code:
@echo off
set SOURCE_FOLDER=full path to the folder with sounds
for /r “%SOURCE_FOLDER%” %%x in (*) do (
set “file=%%x”
setlocal enabledelayedexpansion
set “path=!file:%SOURCE_FOLDER%=!”
adb push “%%x” “/oem/sound/!path!”
endlocal
)
- Change the file extension from .txt to .bat, then run it.
- Done!
To create your own voice pack for Midea, simply prepare a folder with audio files in .mp3 format. Each file must meet the following specifications:
- Bitrate: 64 kbps (audio compression quality)
- Channels: mono (single-channel audio)
- Sample rate: 48000 Hz (number of audio samples per second) File names must correspond to the numbers they represent. More details are in the Midea command mapping file. For convenience, here are the original Midea M7 Pro sound files.
A Xiaomi voice pack is a tar.gz archive containing .mp3 files without folders (all files are placed directly in the root of the archive). All files are named with numbers. Each number corresponds to a specific phrase. See more details in the Xiaomi command mapping. The file bitrate is 96 kbps. The sample rate is 44.1 kHz or 16.0 kHz.
There is a difference in quality between low bitrate (e.g., 32 kbps) and higher bitrate (e.g., 96 or 120 kbps), but for short voice phrases it is almost unnoticeable. Therefore, it is recommended to use a bitrate in the range of 100–120 kbps for an optimal balance between quality and file size.
Unfortunately, I haven’t yet found a way to upload voice packs to xiaomi.vacuums, as I don’t have a xiaomi.vacuum device myself. If you do have one and are willing to help, I’d really appreciate it! You can contact me via email at b.sverdlyuk@ukr.net
Special thanks for the truly monumental work in creating this voice pack to Lost Human.
Thanks to everyone who financially supported this project. Without your donations, it would have been impossible. Thanks to Oleksandr Belei for the Dreame phrase file used in this project.
If something remains unclear or you have difficulties with setup — feel free to ask for help. You can create a topic in the Discussions tab, write in the Smart Home | Matrix chat, or leave a comment in the relevant DOU topic.
🔧 If you notice inaccuracies or have ideas for improvement — I would sincerely appreciate your pull request or feedback.
This project is open and available under the MIT license. You are free to use and modify it according to the license terms.