Currently, watching movies and videos on the internet serves all needs such as learning, entertainment, and research. The application of artificial intelligence in translation and speech recognition is also discussed. They are also developing in many research directions, such as Speech-to-Text recognition applications based on specific audio files. However, studies often focus on improving the processing speed and the accuracy of words converted into text inside the audio file but have not focused on clarifying the voice inside the audio file to facilitate easy and accurate identification. Like no tool can automatically create subtitles for videos for free, but only manually create subtitles based on timestamps and adding subtitles, which is quite time-consuming for long movies or videos. Therefore, this study proposes a new approach by combining audio processing for noise reduction, noise removal, and audio-to-text recognition to create a tool to generate subtitles automatically with high accuracy. The study results are only experimental to create a research direction that can be developed and implemented into viable applications for creating subtitles for videos without having to do it manually and with an accuracy of about 80%.