A new gold rush: It's now possible to build apps that create video from text

In just the last 24 hours, the landscape of AI-generated video creation has dramatically shifted.

Two industry giants in the world of text-to-video announced new APIs:

• Runway announced an API for its Gen-3 Alpha Turbo model.

• Just hours later, Luma, a large Runway competitor, also announced an API for its "Dream Machine model."

This could spark a new gold rush for indie founders.

What does this mean for us as founders?

We've seen hundreds if not thousands of successful new indie SaaS businesses that run on top of text-based AI models like GPT-4.

Many of these reach 4, 5, or even 6 figures a month in revenue, very quickly after launch.

We've even seem some successful indie startups based on image models, like Pieter Levels' PhotoAI app which just passed $100,000/month in revenue.

But there's a gap. Nobody's been able to do any of this for text-to-video AI models, simply because we've never had access to any good APIs.

And these models are prohibitively expensive to train, and thus out of the reach of 99.9% of bootstrapped founders like us to create our own.

So before this new "Dream Machine" API, for example, the only way to make AI-generated videos was one-at-a-time, through Luma's website.

But that all changed this week.

How good are these models?

You probably remember when OpenAI released their famous Sora model, which displayed stunning AI-generated videos.

But then none of us ever got access.

Well, these newly-accessible models are in a similar league. Pieter Levels had glowing praise for Luma's Dream Machine model:

“It's absolutely state of the art and the best video model available right now I think. It's able to highly accurately predict 3d surroundings from just a photo, I think up there with Sora by Open AI.”

Runway's pretty neat, too. Here's a trailer taken from their website, which showcases a few AI-generated videos created with their Gen-3 Alpha Turbo model.

All that said, in my own tests of Luma's Dream Machine model, prompting proved to be difficult. It was usually kind of close to what I wanted, but not quite there. And repeated attempts to tweak my prompts were tedious, because video generation takes at least 20 or 30 seconds, sometimes longer.

How do you get access?

Unfortunately, you need to apply for access to Runway's new API.

Fortunately, the form is pretty simple. Just enter your name, email address, company name, and a brief description of what you plan to use the API for, and they'll get back to you.

It's better to jump on these things sooner rather than later, so you don't get stuck in the back of a long and ever-growing queue.

As for Luma, you can start using it immediately, today! Their web UI is free, but pricing for the API is a bit complicated:

• $0.0032 per 1M generated pixels

• …which is about $0.0032 per frame at a resolution of 1280×720p

• …so about $0.40 per 5 seconds of video at 24 frames per second

What are some business ideas you can build?

Danny Postma recently tweeted his formula for "how to get rich in the age of AI." And he should know, because he's done it:

So just to get your creative juices flowing, here are a few ideas for software tools you could build on top of these text-to-video APIs, the first of which follow from Danny's advice.

1. Online Course Content

Target online course creators. Create a platform that will automatically generate instructional videos directly from their lesson outlines, which should allow them to develop fully-featured courses much faster and cheaper.

2. Corporate Training Videos

HR teams inside big companies often spend weeks coordinating with video production agencies to create training videos. Imagine if instead they could simply type a description of the training topic and instantly get a professional-quality training video to help employees with onboarding, compliance, skill development, etc.

3. Content repurposing

Allow users to enter the URLs for a collection of blog posts or articles, then use AI to extract the most important parts of the content, use a text-to-speech API like Eleven Labs to narrate it, and use these text-to-video APIs to generate a relevant video. You could pretty quickly turn a catalog of articles into a library of videos.

4. Marketing video ad generators

Build a tool where people can input their business goals and the video ad formats they want, e.g. Instagram, Facebook, etc. Automatically generate videos for them to use in their ad campaigns.

5. Personal avatars and profile photos

From Telegram to Tinder, lots of apps allow users to upload videos to represent themselves in their avatars and profile photos. Both Luma and Runway support using photos as a reference for creating video files. You could generate a whole host of fun, attractive, or quirky videos for your users to represent themselves on other apps.

6. Motivational quotes

Video might be the missing piece for creating motivational quotes that are actually engaging. Currently there are tools that can generate realistic voices, but most of the videos behind those voices tend to be bland.