Inspiration
Assuming that advertising revenue is a reasonable proxy for attention, it turns out that humans like pictures more than text, and moving pictures most of all; so it has gone on the Internet.[source].
What it does
This bot takes the URL of a web-based content article and automatically generates a motion-video out of it.
How I built it
Newzery uses moviepy for making a video out of text, images, and sound. The article of content is scrapped using newspaper. The sound effects are determined by the sentiment analysis of the content, which is done using nltk. The bot is made using flask. Facebook messaging API is integrated with the bot using pymessenger. The bot is hosted on Heroku free tier.
Challenges I ran into
- Compression of video to match the attachment file limit of messenger API
- Handling requests to avoid exceeding of memory quota on Heroku free tier
- Setting up the production environment for moviepy: installing ImageMagick on Heroku
Accomplishments that I'm proud of
- automated delivery of personalized media over messenger
- enhanced retention of article content in users' memory
- a major step towards an end-to-end pipeline for on-demand article2video generation
What I learned
- messenger webhook callbacks
- garbage collection in python
- meeting production server constraints
What's next for Newzery
- Iterations for getting the product market-ready
- Expand to more platforms including Whatsapp, Instagram


Log in or sign up for Devpost to join the conversation.