Last Updated on
At the dawn of generative AI, one winner will emerge from each school of digital media; Text, audio, image, and video. Long ago, the four nations lived in harmony. ATLA references aside, RunwayML has emerged as the champion of generative video AI, and in this article we answer the questions “What is Runway ML and what does it do?”
What is Runway ML?
Runway is an independent AI research lab by the tagline “No lights. No camera. All action.” in reference to their ambitions of disrupting the video production industry. At it’s core, it’s software. In more reasonable detail, it’s a platform of machine learning models; An artificial intelligence specifically designed to produce original, aesthetically-pleasing and realistic (if required) video from simple text prompts. In their own words, their software will “Realistically and consistently synthesize new videos with nothing but text. It’s like filming something new, without filming anything at all.”
Essential AI Tools
Winston AI detector
Originality AI detector
Producing high quality video has always had an equally high barrier of entry. The big-budget world of Hollywood notoriously gate-kept and inaccessible. Even before that, the field of videography comes with a steep learning curve from the moment you type “How do DSLR’s work” into Google.
The vision is to put the power of video into the hands of… everyone. Without the cost of a camera, or the logistics of travelling to your shoot location, or the years of practice it takes creatives to consistently make good decisions about angles, lighting, and colour. The mark of a powerful AI is how much of the learning curve it removes. Using AI should feel like cheating. If using AI doesn’t make things easier, when what was the point in making it?
What does Runway ML do?
Type a text prompt into Runways interface and it will export videos that look like whatever you described. Boasting a next-generation creative suite of ml models, with new models like Gen-2 added periodically, creators can produce original video clips using nothing but internet access and an idea. Their suite of what they call “magic tools” bypasses the video editing / vfx process, and encompasses the following generative models:
- Gen-1 Video to Video
- Gen-2 Text-to-Video
- Frame interpolation (Separate models for smooth slow motion images-to-video)
- Erase and replace (Using mask brushes)
- Infinite Image (Out painting)
- Colour grading
- Audio noise removal
- Green screen tracking
Runway is more than your new video editor, it’s a complete AI-powered video production studio, and one of the most exciting start-ups in the space right now.
Who owns Runway ML?
Considering their rise to the forefront of an industry with little (if any) attention on the people behind it, you might ask “Who is the founder of RunwayML?”. Cristóbal Valenzuela is the CEO and Founder of Runway. Formerly a researcher at New York University, Cristóbal began work on his own venture in December 2018 and has been hard at work building AI for almost 5 years since.
How much does Runway AI cost?
Based on the value this will bring to many industries and professional individuals, you might ask “Is Runway free to use?”. There is a “Free Forever” Basic tier, which I find incredible considering the inherent costs of an AI enterprise.
The free version includes 125 credits per month, roughly translating to 25 seconds worth of generated video. On this tier, exports will be watermarked and will only export in 720p. It’s enough to check that it does what you want, but nothing you could professionally use.
Higher paid tiers will of course remove the watermark, allow usable quality exports in a wider variety of file formats, more cloud-based server storage to back up more simultaneous projects, and most excitingly the ability to train your own custom AI generators. This means applying Runway algorithms to your own dataset of images (training data) and video to substantially influence the results you consistently produce.