MoodSmith 🕊️

Mood-Consistent Content Creation for Non-Profit Organizations 


Many organizations, despite having established messages, face challenges in creating advocacy campaign videos for social media. It requires significant resources and time to develop meaningful content to express these messages while achieving a consistent mood across multiple dimensions: script, visuals, and audio.

MoodSmith helps you draft short advocacy campaign videos that communicate your ideas in a cohesive narrative through text, imagery, and music.

Interested in trying it out? Contact sm4788@columbia.edu.

Example Message: Every year, around 200,000 birds lose their lives due to collisions with buildings in NYC alone. Advocate for future construction to use bird-safe glass.

Mood:

Delighted

Depressed

How do you use MoodSmith? ✏️

This tool has 4 stages: the Initial Input, Script-Editing, Storyboarding, and Video Preview. These stages allow you to brainstorm and edit along with the system to ensure that the video produced meets your vision and requirements.

  • In this stage, you input the audience, the mood, the problem/topic you're addressing, and information or solutions you would like to share. If you have any images you'd like to incorporate, upload them here too.

  • In this stage, you edit a 'script' of multiple scenes. Each scene includes a narrative goal, on-screen text, an image description, and an analysis of how "positive" the scene is. You can edit, regenerate, and tweak the narrative goal and positivity score to fine-tune the scenes.

  • You're almost done! Here you'll get to see a storyboard of the scenes you edited earlier, with images, onscreen text, and the positivity analysis from earlier. Here, you can continue editing and regenerating the images and text.

  • Based on the mood of your video, the system will have a few music recommendations to help you explore what type of audio you would like on your video.

How does it work? ⚙️

MoodSmith employs various tool to support the video creation process, including GPT-4, Stable Diffusion, and the Spotify Web API. For more information on how our system works, please check out our paper here.