How to Make an AI-Generated Film 🎬


Hello Reader,

Guess what? I made my first AI-generated film!

🎬 Making an AI-Generated Film

I was thrilled to participate in RunwayML’s GEN:48 AI Video Short Film competition with my former coach and mentor, Thomas Haynes.

The competition was exciting - we didn’t know the topic until 9 a.m. Saturday morning, and we only had 48 hours to complete it.

We had to choose at least one element from three categories:

  • Character Archetype: a crime boss
  • Scene: an abandoned building
  • Object: an old corded telephone

We’re enrolled in an AI Filmmaking course by Curious Refuge, and we used some of the tools and techniques we’re currently learning about.

Workflow:

We used Curious Refuge’s AI Filmmaking Pipeline checklist to organize our project.

  • ChatGPT: generate plot ideas, script, and shot creation
  • Google Docs & Sheets: refined script and organized shot list
  • Midjourney: generated a small number of images
  • RunwayML Gen-2: generated 4 to 16-second video clips
  • Pixop: enhanced video clip quality
  • Epidemic Sounds: selected sound effects and music
  • ElevenLabs: generated voice narration with text prompts
  • WeTransfer: transferred videos and images to collaborate
  • CapCut: edited video and combined all media

The short film was approximately 95% text prompt to video-generated with Runway’s Gen-2 app.

This is quite different than prompting individual images for animation. It forced us to think about each scene as a part of the overall story, including the setting, actions, and mood.

It’s a long, meticulous process, and it was worth it. And Thomas is new to AI art and video creation - he just started generating images about 2 months ago!

Watch our film here and let me know what you think:


I recently sat down with Jeff Wilser to discuss my journey to generative AI and AI Art.

Listen to my Apple Podcast interview on AI-Curious with Jeff Wilser here.


🔮AI News This Week

Midjourney Upscale

Midjourney added two Upscale options for 2x and 4x.

You'll see the new Upscale buttons under all new upscales in v5 and Niji, and after you've upscaled, you will see "Redo Upscale" buttons below the image for 2x and 4x:

Things to consider with the new Upscale feature:

  • Only work with Fast GPUs
  • 4x upscale uses more GPU minutes
  • Buttons visible under normal v5 & Niji upscale
  • Possible darkening, black, or corrupted images occur rarely

Note: The file size is much larger than a normal Midjourney upscale! Normal upscaled images were less than 2 MB. The 4x upscale will be around 20 MB or more.

Consider if you really need to use these options to preserve Fast hours and minimize disruptions to your workflow.

Here are examples of my original upscaled image, compared to the 2x and 4x upscaled versions:

Alternative Upscale Tools:

It’s convenient to have an upscale feature available while you’re generating Midjourney images in Discord, but it might not meet your expectations, or you can upscale later if you choose.

Many people already have a workflow in place with other tools. Either way, plenty of tools are available for any upscaling and enhancement.

  • BigJPG: Free image upscale tool uses neural networks and “intelligently reduces noise and serration in images.” Additional tools available include AI Image Magic Cleanup, AI Video Enhancement, and AI Photo Restoration. (link)
  • I💙IMG: Free site with a ton of tools, including image upscaling, file compression and conversion, and watermarking. (link)
  • Upscale.media: Free image upscale & enhance tool, supports most file formats, including webp. (link)
  • IMG Upscaler: Free bulk image upscaling, background & object eraser. (link) Check out my post with demos of the object eraser, face-swapping, and background remover here on 𝕏.
  • Upscayl: Open-source upscale tool available for MacOS, Windows 8 and above, and Linux, with cloud services coming soon. Join the beta here: (link)
  • Topaz: Photo and Video upscale tools you purchase with a one-time license fee. (link)
  • Pixop: Video upscaling, restoration, and enhancement right in your web browser. Pay as you go. I use this frequently to enhance my AI-generated video clips from Runway Gen-2 and Pika because it’s fast and doesn’t require an app download. (link)
  • Pixelcut: Image upscaling and enhancement, plus several editing tools and a profile picture maker. (link)

Niji•Journey App

Niji•Journey finally has its own app, available for browser and mobile. It has only been available in Midjourney’s Discord, and Midjourney will be launching its web app in the coming months. (link)


You could have your AI service, tool, or event seen by Visually AI’s community of over 6,300 subscribers: Advertise with me


🚀 This Week’s AI Tools

Tweet Hunter: 𝕏 and LinkedIn content scheduling tool with several GPT-4-powered AI tools, including the new Video Signature feature. I’ve been using Tweet Hunter for over a year, and it’s essential for me to schedule my content, refine text, and engage with followers and other accounts. (link)

WeTransfer: Free file transfer up to 2 GB at a time. Send large files, including images and videos, without losing quality via email or a link to download. (link)

Adobe Express Image Resizer: Fast, free online image resizing tool. (link)


🎁 Get it free: The AI Visual Creator’s Toolkit

Boost your content with my all-in-one, free visual AI toolkit!

Access AI-powered tools for AI-generated images, image editing, and more: Get your toolkit


Visually AI

Your weekly dose of AI news, tools, and innovation with a visual twist. Breaking down barriers to make AI content creation accessible to all. Expressing my own personal takes and educating readers in the world of Artificial Intelligence.

Read more from Visually AI

Generated in Midjourney Hello Reader! Welcome to Visually AI! Publish & Remix Claude Artifacts This week, our focus is on Anthropic’s Claude. Anthropic’s Claude is designed to handle complex natural language understanding tasks, offering capabilities such as summarization, question answering, and content generation. Anthropic has introduced a new feature allowing Claude Artifacts to be publicly published, enabling others to view and remix these creations. This functionality promotes...

Midjourney Year One Contents Hello Reader! Welcome to Visually AI! Thank you Midjourney! Yesterday, was like every other day. Working hard (recording a podcast!) and bustling away on the multiple projects I am doing. And then I was tagged by Tom Blake. Something about being included in Midjourney's Year One Book... So I clicked on it... And there I was: I was overwhelmed with emotion. It's been an incredible journey so far into Artificial Intelligence. From collaborating with Inter Milan,...

Generated in Midjourney Hello Reader! Welcome to Visually AI! Where to begin? Shall we start with OpenAI who bought a couple of companies? What about ChatGPT being widely available on Mac? Or the fact Toys 'R' Us made an ad with Sora... yes SORA! The AI video generation tool we've all been waiting for. Or perhaps Eleven Labs partnering with Time. How about Claude 3.5 Sonnet model? Hmmm, maybe Suno and Udio getting sued by labels? Or an AI Al Michaels announcing the Olympics... Okay... Let's...