Skip to content

DominiquePaul/dreamstream

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

94 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

DreamStream - AI-stylised Video Streaming Platform

Website GitHub Repo

DreamStream is a cutting-edge video streaming platform that leverages AI technology to stylize live video streams in real-time. This innovative platform allows users to create and share unique, AI-generated video content, revolutionizing the way we interact with live streaming.

Check it out at www.thedreamstream.io!

Homepage Screenshot

1-Week Development Sprint Timeline

The following timeline outlines the key milestones achieved during our 1-week development sprint:

Date Goals Achieved
Monday (a) Researched img2img technology and selected a suitable network for implementation ✅
(b) Developed a basic class for style transfer ✅
Tuesday (a) Designed the infrastructure for live streaming ✅
(b) Created a minimal website with streamer and viewer interfaces ✅
(c) Deployed the website to a live URL and set up continuous deployment ✅
(d) Deployed a basic WebSocket server ✅
(e) Enabled users to create and view live streams via the website ✅
Wednesday ...
Thursday ...
Friday (a) Deployed the WebSocket server on modal.com for autoscaling capabilities ✅
(b) Integrated image transformations using H100 GPU on modal.com ✅
Saturday (a) Removed black patches between frames in the stream ✅
(b) Allowed users to specify prompts for image transformations ✅
(c) Optimized the speed of image transformations ✅
Sunday (a) Integrated lightning diffusion code into the modal.com deployment ✅
(b) Implemented registration and login for streamers ✅
(c) Enabled streamers to purchase credits for streaming via Stripe checkout ✅
(d) Added real screenshots and enhanced the app's UI ✅

🚀 Features Roadmap

Upcoming Features

  • Implement real-time chat functionality for each stream
  • Display the number of viewers for each stream to streamers
  • Users can request style and streamer can accept/decline OR streamer can set auto-accept
  • Streamers can set a theme, e.g. "90s TV shows" and set the prompt to change every 120 seconds.
  • Streamers can export all frames/timelapse of all frames after download
  • Better SEO

Future Development Ideas

  • Develop an admin dashboard for seeing usage stats
  • Introduce audio streaming capabilities
  • Enable live streaming from raspberry pi devices

Long-term Vision

  • Allow users to utilize the platform as middleware for streaming to other platforms (e.g., YouTube, Twitch)
  • Explore the possibility of supporting multiple camera streams simultaneously, enabling unique streaming effects not achievable with a single camera

Product Showcase

Here's a glimpse of what DreamStream can do:

Stream Example OriginalAdjusts to the structure of people and movements in your videoStream Example Augmented

Collection of things I want to take a closer look at to improve the model speed

About

Livestream from your home or bar with styletransfer applied automatically

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors