Text-to-video using Runway Gen 2 | Demo

Introduction

If you're a video creator or editor, this piece of news will excite you: the highly anticipated upgrade to Runway ML, Gen 2, is here. The previous AI model, Gen 1, allowed users to transform existing videos into new ones through image references, presets, or text prompts. Now, with Gen 2, you can make videos from scratch by providing text prompts alone.

Accessing Runway Gen 2 is fairly simple. All you have to do is type in app.runwayml.com on your web browser and create a new account or log into your existing account. On the homepage, click on the "Introducing Gen 2" tab, which will take you to the prompt screen. You can also access Gen 2 by clicking on the "Generate Videos" option on the left menu bar and selecting the "Gen 2 Text to Video" option.

In the prompt space, the filter icon on the bottom left allows you to interpolate videos to make them smoother, upscale the videos, and remove watermarks. The Gen 2 tool can currently generate 4 seconds of video for free accounts; for paid accounts, the duration of a single video can be up to 15 seconds.

On the right side of the prompt space, you can add images and type in the video requirements in connection with the image. Let's try some prompts out. It's important to note that when we tried a few generic prompts such as "cat in a garden" or "a quiet street," the tool came up with moderately accurate videos. The more descriptive we got, especially in terms of style, the less satisfactory the video results were.

We found videos on the internet that prove otherwise. A better understanding of camera and editing features like depth of field and long exposure can make your prompts deliver the results you expect. It's also challenging to generate videos with specific styles, genres, or artists. You can prompt for "Da Vinci style" or "cyberpunk style," but there are no significant or visible changes in the video results.

Runway has always positioned itself as a beginner's tool, aiming to help filmmakers who are just starting out. For example, features like background removal or noise reduction can be done without mastering the Adobe Creative Suite. The text-to-video features align with that ideology. The product isn't yet refined, but its target audience isn't your After Effects expert. The bottom line is that Gen 2 is still nascent, and like any new tool, it requires quite a bit of work. However, it does look promising, so let's hope it gets better.

Thank you.