"LoFi Focus team on AudioCraft 24-hours Hackathon Hackathon"

LoFi Focus

## Implementation - Built as a Chrome browser extension for ease of use - Uses JavaScript content scripts to analyze webpages and play lofi audio - Leverages AudioCraft's MusicGen AI model to generate the lofi tracks - Polished UI allows easy control over the music generation --- ## Our Custom Model We collected a dataset of original non-copyright lofi music. This gave us access to a large corpus of high-quality training data without any copyright issues. We split the lofi songs into 30 second audio clips and paired each clip with a text prompt describing the mood, instruments, tempo and other qualities of that segment. Examples include "slow chill hip hop beat with mellow piano and vinyl crackle" and "upbeat lofi with energetic drums and warm bassline". We formatted this dataset into the required .wav and .txt file pairs that musicgen_trainer expects. The text prompts would guide the model to learn the nuances of lofi hip hop. We then ran musicgen_trainer on this dataset, configuring it to use the small architecture for optimization purposes. We trained for 100 epochs with a learning rate of 1e-5 and batch size of 4. During training, musicgen_trainer used the audio/text pairs to fine-tune MusicGen on lofi music. The pre-trained weights were specialized to generate high quality lofi given descriptive prompts. After training finished, we saved the best performing model checkpoint. We now have a MusicGen variant skilled at generating original lofi tunes according to textual descriptions. --- ## Why Download Our Chrome Extension - Improve focus and concentration when reading - Make reading more enjoyable and relaxing - Boost productivity - Avoid listening fatigue - Portability - Ease of use - Less anxiety - Nostalgia