Table of contents.

1. Author disclaimer
2. Introduction
3. State of the art
4. Requisites
5. Project structure
6. Installation
7. Configuration
8. How to use
9. Limitations and considerations
10. Future work
11. Demo
12. References
13. Licensing agreement

This project has been carried out solely and exclusively for self-learning reasons and to show it as a career portfolio.

I am not, and I will not be responsible for any type of misuse with the code of this project by third parties. At no time has it been my intention, nor will it be to cause any kind of harm to Tinder and its products.

You will surely wonder why I have decided to do such a project and if it is appropriate to have it in my GitHub account for companies to see... the short answer is: have fun.

Outside the workplace, I like to program things that I have fun with while learning new tools to improve my skills. If I don't have fun learning in my spare time, I don't program.

Most of the time YouTube recommendations are useless but in some rare cases you can find something interesting. That is what happened to me when I saw a video of Code Drip about making a bot for Tinder and after seeing that he only contemplated the basic cases, I came up with the idea of making an improved and more complete version.

In addition, for a long time and due to the current pace of life for many people because of their work, there are people who hire a person to manage their dating applications for them. Here is a well-known case: How one woman turned writing dating app messages for busy professionals into a full-time job and there are agencies that offer this kind of service, and their prices are not cheap.

Then I thought, why pay for others to manage your dating applications when a few lines of code and a neural network can do the same thing for free?

Well, that is the reason why you are reading this.

Dating applications are a very juicy niche due to the large volume of benefits they produce. From time to time a new dating application appears that, although it promises to be something different from the rest and offer something new, in the end it is still more of the same. Because of benefits they generate, many people have taken advantage of this to implement bots for Tinder but the problem with these bots is that they are very simple and only address basic base cases.

When performing a quick search through GitHub, you can find repositories with old and obsolete bot implementations, in other cases they work, but they only use a random number to decide whether to click on like or not.

On the other hand, there are some projects that have managed to obtain the Tinder API. While they are not always updated to the latest available version, often using that method ends up with prohibited accounts, including the phone number associated with the account. This is not a desirable option.

Therefore, the implementation that has been carried out in this project seeks to simulate how a human would interact with the application and, to achieve this, the web version that the application provides to users is used. To cover a wider range of users, the bot has been designed for a user who has a free account with a limit of daily profiles that he/she may like although users with a paid version can also use it, but it has not been tested with that type of account.

Due to the limitation of the number of times a user with a free account can like a profile, the bot can optimise the number of times he/she likes a profile by being able to discard profiles without photos or profiles with photos in which no human face appears. More realism has been added giving the bot the ability to predict the beauty of a profile, checking the symmetry of the face of the person in question and comparing the average score with the cutoff threshold set by the user. This is achieved using neural networks that are capable of detecting human faces and calculating facial symmetries.

Finally, the bot has become even more humanised by randomly waiting a few seconds between operations in order to go unnoticed between the bot detection measures. Making it appear that the bot is a human who is using the profile avoids overloading the system with multiple requests per second and thus the bot makes a respectful use of the service without affecting the other users who are using the application at the same time.

In order to use the bot it is necessary to have a compatible environment:

• Operative system: Linux or Windows (The crontab functionality has been implemented with Linux systems in mind, not Windows).
• Python version: The minimum and necessary version is 3.5 (due to dependence with Path) although the desirable version would be 3.7 (due the insertion-order preservation of dict objects).
• Browser: Firefox version >= 60. This bot has been designed to use the GeckoDriver and in the latest version they increased the minimum version of the browser to the version indicated at the beginning.
  • Be sure the tracking protection is not enabled on the Tinder website, or the login button with Facebook will not work.
• Tinder: It is necessary to have an active and valid account to use the bot.
  • The account must be created through the Facebook login, otherwise the bot will not work.
  • You must unlock all the buttons in the selection panel. Slide some profiles until the rewind button, the super like button, and the boost button are unlocked for the account. Otherwise, the bot cannot find the correct position of the like and not like buttons.

In this section you can have a quick view of the project structure.

├── beauty
│   ├── model_human_face_detector.dat
│   └── model-ldl-resnet.h5
├── gold_matches (*)
├── logger.py
├── logs (*)
│   ├── critical.txt
│   ├── debug.txt
│   ├── error.txt
│   ├── information.txt
│   └── warning.txt
├── matched_photos (*)
├── predict.py
├── README.md
├── requirements.txt
├── resources
│   ├── constants.py
│   ├── firefox_xpaths.py
│   └── log_configuration.yaml
├── slipped_profiles (*)
└── tinder_bot.py

Directories marked with a (*) will be created by the bot as needed.

This section expects the requirements stated in the previous section to be met and this is how this section has been written.

• GeckoDriver: The driver that Selenium uses to interact with Firefox.
  • Download the file from the GitHub repository: Link
  • Place it in /usr/bin or /usr/local/bin directory (you may need administrator permission).
• Beauty predictor: The predictor needs two files to be able to work.
  • Face recognition model: This model is the one used to recognize faces in photos. (Link)
  • Face beauty model: This model is the one used to measure the beauty of a face. (Link)
  • Place both files in the beauty folder of the project and change their names to model_human_face_detector.dat and model-ldl-resnet.h5 respectively.
• Program dependencies: The bot has some dependencies that must be installed in order to work. Those dependencies can be installed with the requirements.txt file:
  • pip install -r requirements.txt

Note: If you have both Python 2 and Python 3 installed on your system, use pip3 instead of pip.

There are some parameters that need to be set by the user, so the bot can work. Those parameters are in the /resources/constants.py file.

• EXPAND_BUTTON_TEXT: Tinder uses the same button to show more login options as to fix login issues. The text MORE OPTIONS must be indicated in the same language in which the browser to be used is configured. Check on the login screen how it is written, it is case-sensitive.
• SENT_TEXT_CHAT: The text that Tinder displays in the chat when a message has been sent. The text Sent must be indicated in the same language in which the Tinder account to be used is configured. Check how it is written on the chat screen, it is case-sensitive.
• READ_TEXT_CHAT (For paid account profiles only): The text that Tinder displays in the chat when a message has been read by the recipient. The text Read must be indicated in the same language in which the Tinder account to be used is configured. Check how it is written on the chat screen, it is case-sensitive.
• WAIT_MATCH_ANSWER: Boolean value to specify whether the bot waits until the profile has responded or not before sending a message. It is recommend to set the value to True instead of False to avoid flooding matching profiles with messages.
• SILENT_MODE: Boolean value to specify whether the bot starts the web browser with an interface or not. It is recommended to set the value to True instead of False to avoid a greater consumption of resources.
• DEFAULT_CHAT_MESSAGES: The list of default messages that the bot will select when sending messages to the matching profiles.
• SCORE_THRESHOLD: The beauty threshold that the bot will take into account to decide whether to like or dislike a profile.
• USERNAME: The email associated with your Facebook account.
• PASSWORD: The password associated with your Facebook account.

There are other constants in the file, but they should not be changed unless you know what you are doing or if you want to continue with the development of the bot.

It is highly recommended to use a virtual environment (venv), so the bot dependencies installation will not conflict with the packages installed on the system. By default, the crontab functionality expects to run in a virtual environment (venv), so the path of the virtual environment must be specified in the crontab_script.sh file:

#!/bin/bash

source path_to_your_virtual_environment/bin/activate
python3 tinder_bot.py

If, on the other hand, you want to run the bot without a virtual environment (venv), the source line must be commented.

#!/bin/bash

# source path_to_your_virtual_environment/bin/activate
python3 tinder_bot.py

The idea of this project is to implement the bot on a Raspberry Pi and leave it running without seeing what is it doing (although, it is recommended to see the log files occasionally to check that it is working as expected).

Since facial recognition and beauty prediction are time-consuming, it takes the bot about 120 minutes (2 hours) to reach the daily limit of 100 likes and the time also depends on the number of profiles discarded when they are invalid. The fewer profiles the bot has discarded, the less time the process will take.

To start the bot, you must grant it execution permissions or, otherwise, it cannot be executed. In addition, it also needs the permissions to be able to generate an entry on the crontab file (Linux systems only):

chmod ugo+x tinder_bot.py

If you are going to use the crontab functionality, you must also grant permissions to the bash script:

chmod ugo+x crontab_script.sh

After that, you can start the bot with:

python3 tinder_bot.py

You can decide what the bot can do by commenting (adding an # at the beginning of the line) or uncommenting (removing the # at the beginning of the line) some function calls:

bot = TinderBot() [1]
if bot.check_user_constants(): [2]
    bot.login() [3]
    bot.close_permission_popups() [4]
    bot.auto_swipe() [5]
    bot.collect_photos_matched_profiles() [6]
    bot.review_messages_from_matching_profiles() [7]
    bot.check_blurry_matches() [8] # DEPRECATED: It no longer works.
    bot.remove_photo_folder(bot.temporary_folder) [9]
    bot.check_crontab_entry() [10]
    bot.exit() [11]

1. Initialize the bot. (Mandatory)
2. Verifies that the values ​​entered by the user are of the correct type. (Optional, but recommended)
3. Sign in to Tinder through Facebook. (Mandatory)
4. Closes the Tinder pop-ups to request permission to detect the location, be able to send notifications and the cookie banner. (Mandatory)
5. The bot tries to find your soulmate. (Optional, although I'm not sure...)
6. The bot downloads the photos of the profiles with which you have had a match. (Optional)
   • This feature has a parameter, skip_animation, which is used to control when it is executed. If the value is True it means that it runs after step 5 and if the value is False it means that it runs without step 5. This is important because otherwise it will fail.
7. The bot scrolls through the list of your matching profiles and sends a predefined message chosen at random. (Optional)
   • To use this feature, you must have previously entered some values in the message list.
     • By default, the bot will not send a message if the matching profile has not yet answered. This behaviour it is controlled with the constant WAIT_MATCH_ANSWER in the file /resources/constants.py.
     • Remember to always be respectful to everyone.
     • Modify the line DEFAULT_CHAT_MESSAGES = [] in the file /resources/constants.py with some values, for example DEFAULT_CHAT_MESSAGES = ["Hi!", "How are you?", "How's your day going?"].
8. [DEPRECATED] The bot gets the photo of the blurred profiles if the user has a free account. (Optional)
   • This option has not been tested on a premium profile.
   • This functionality has stopped working since May 1, 2020. Now it seems that Tinder checks if your account has permissions to access the original photo of the blurred thumbnail, so when the bot tries to obtain the original photo you get an access denied error instead of the original photo.
9. Deletes the temporary folder where all the profile photos have been downloaded. (Optional)
10. Sets the bot to start automatically after turning it off. (Optional)
    • This function only works on Linux systems.
    • If the bot has turned itself off because there are no more profiles to display, it sets itself to start again in an hour.
    • If the bot has turned itself off because the daily limit on the free account has been reached, it sets itself to start again in 12 hours and 5 minutes. Although the Tinder cooldown lasts 12 hours, the bot gives 5 minutes of margin.
11. Turns off the bot. (Mandatory)
    • The bot will also turn itself off without this call but the driver instance will not be deleted. The browser window will remain open, consuming memory resources if the SILENT_MODE it is not activated.

Note 1: All the photos downloaded, except for gold_matches, will be saved in their respective folder inside a folder called profile_name - profile_age - number_of_photos.

Note 2: The bot will automatically delete the gold_matches and matched_photos folders when you have no more matches.

• The bot has certain technical limitations that can cause some profiles that are valid from the user's point of view to be discarded.

  • If the photo has a partial face (it is cut off) the bot will fail in the beauty prediction phase as it cannot obtain the necessary points to perform the calculation, so that photo will be invalidated with a score of 0.
  • Although a face can be detected in the photo in the pre-beauty prediction phase, the photo can be discarded if the person appears too far away in the photo, and the face is not large enough to get all the points. In this case, the photo will also be invalidated with a score of 0.
  • Sometimes some profiles have pictures of drawings with human faces. If the drawing is very simple, the bot detects that it is not human, but if it is a realistic drawing, the bot can be confused and identify the photo as valid.

• There are some considerations to keep in mind when calculating the beauty score:

  • When calculating the beauty score, keep in mind that the same value will not always be obtained for the same person. The value obtained depends on the quality of the photo, the amount of light in the photo, and the expression of the face that appears in the photo. The score will also be affected if there are filters from Instagram or another application in the photo as they confuse the neural network.
  • Another aspect to consider when calculating the beauty of a profile is the extreme values. When analyzing profile photos, three options can happen in each photo if the person appears with more people because the beauty predictor returns the scores of all the faces found in each photo:
    • If the other people in the photo get a score similar to the person of the profile, those values ​​do not affect since it would be as if the owner of the profile has more photos uploaded.
    • If someone in the photo gets a much higher score than the person of the profile, that score will positively benefit him/her.
    • If someone in the photo gets a much lower score than the person of the profile, that score will negatively penalize him/her.
    • After consulting with several statistical friends, the easiest way to avoid these deviations is to use the median, since it ignores the extreme values of the intervals.

At the time of writing this document and uploading the code, the bot works completely. This does not imply that this will be the case in the future as the web version of Tinder experiences changes and consequently the buttons and sections paths change.

Chatting functionality could also be improved. There is a specific case in which the bot can consider that the person has not responded if the last message the person sent is Read or Sent (in the language in which the Tinder account that uses the bot is configured).

Therefore, if at any time the bot stops working, the xpaths should be reviewed and updated.

At first, I had thought of putting out a video with a demonstration of how the bot works but because I was not sure if I could show the profiles of people without their consent despite the fact they are exposed in a dating application (although on YouTube you can find videos where people show profiles) finally I decided not to release the video, so if you want to see the bot working you will have to install it and run it locally on your computer.

• Beauty predict project by ustcqidi: GitHub
• Tinder bot example project by aj-4: GitHub

Copyright © 2020 MinionAttack

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.