AI/ML Engineer - Vibecoding some side-projects

Juan Piñeros  • Peerlist

Video Moment Finder (Beta) offers an innovative solution for locating specific moments within videos using AI-powered semantic search. This tool allows users to find moments using text or images, making video navigation more intuitive and efficient. Key features include the ability to search through video content semantically, leveraging advanced AI technologies to understand and interpret video frames. This open-source project is licensed under AGPLv3, ensuring transparency and community collaboration. The source code is available on GitHub, inviting developers to contribute and enhance the tool further. Whether you're a content creator, marketer, or developer, Video Moment Finder provides a robust platform for video analysis and management.

Video Moment Finder

By the end of this project, you will create a music web application that uses AI to generate melodies,  based on the musical idea given by the user. For this, we are going to use Magenta.js, Vexflow and Tone.js. In order to take this project, you should be familiar with HTML, CSS, and JavaScript. Starting from scratch, we will use Glitch to create our music web app. If you are interested in music and machine learning, this project is for you. This Guided Project was created by a Coursera community member.

Create a web app that generates melodies usin

Primera red de Latino América que promueve procesos de E-Mentoring

En Busca de Sentido www.enbuscadesentido.com

Music game to use melodies generated by Magenta in an intuitive way. I used P5.js Tone.js and Magenta.js to create the demo as a part of the "BitRate: Machine Learning & Music Series".

Eternal Circles

This code was built on top of the implementation of Dr. Tristan Behrens of the "MMM: Exploring Conditional Multi-Track Music Generation with the Transformer" paper. The changes to the original code are just a few adaptations so the repo could work with the LMD dataset.

https://github.com/juancopi81/mmm_tokenizer_l

Komposair is a web app that lets musician create melodies using A - Create melodies based on your motif - Explore other melodies created by the community - Save your melodies, download them as midi files, vote for your favorites and comment them In the browser Komposair needs no installation you can use it right in your browser. We use Tone.js and Magenta so you can hear the melodies in the browser or export them as a midi file if you prefer You can also: - Upload your model (to be done, for the moment feel free to send us and email with your model) - Find a random melody if you feel lucky The models to create melodies will be improved, the melodies are generated using motifs of the users. The motif is one building block for melody compositions. Arnold Schönberg defines a motif as: "a unit which contains one or more features of interval and rhythm [whose] presence is maintained in constant use throughout a piece"

Komposair

Community pipeline line merged in the official Diffusers library that enables to use Stable Diffusion in 50 languages.

Multilingual Stable Diffusion

This Stable Diffusion model has been fine-tuned from Riffusion to generate spectrograms of Currulao music. Currulao is a traditional Afro-Colombian music and dance genre, characterized by its rhythmic beats, call-and-response singing, and lively percussion instruments, that holds significant cultural and social importance in Colombia, particularly in the Pacific coast region, as a celebration of African heritage and community identity. To generate the concept, use the phrase ‘a $currulao song’ in your prompt.

Riffusion-Currulao

Tandemia is a Tinder-Like app that helps you find someone near you to exchange languages. Talk with native speakers. Filter your search by language, distance, age and gender.

Tandemia tandemia.co

Transcribe music from YouTube using the MT3 model by Magenta Google.

YouTube Music Transcriber

Create a video-summary of a YouTube video. Supports translation. Project selected as one of the winners in the Creative Challenge of the Whisper event supported by Hugging Face and Lambda Labs