What About Adobe’s Newly Released AI Video Model?

Adobe AI

Adobe AI
Adobe HQ, Credit: Coolcaesar – CC BY-SA 3.0 via Wikimedia commons

As the AI “wars” rage on, companies are actively trying to outdo each other, and Adobe has no plans on falling behind competitors. Keeping up is key to tech and software companies if they are to survive. Hence, the company has now released its Firefly Video Model.

The new AI tool is already available across its products, such as Premiere Pro, allowing users to extend footage and generate video from images and text prompts. Essentially, it works like ChatGPT but for video.

The first tool is Generative Extend, launched as a beta for Adobe Premiere Pro. It lets the user extend footage to their liking and also enables them to make adjustments mid-shot.

Adobe AI’s limitations

However, there is a reason why Adobe Firefly is still in beta, as it has some limitations.

For instance, clips can only be extended by two seconds. This AI is still in its infancy and is meant to be used for small tweaks, at least for now.

Extended clips can also be generated at 720p or 1080p at 24 frames per second. Additionally, Adobe’s AI can also be used for audio, but it can only do it for 10 seconds and doesn’t work on dialog or music.

Adobe seems to be putting all of its eggs in the AI basket, as it is releasing two video generation tools. It is currently also testing the Firefly web app as a limited public beta.

Moreover, Adobe’s Firefly can emulate several styles of video. The AI can generate real-style frames, 3D animation, and stop-motion with fully customizable camera angles and shooting distance.

The AI cannot produce long-form videos

An important point of contention in generative AI debates has been that the technology might replace creatives in the future. This was one of the main talking points during the SAG-AFTRA strike and will likely continue to be for the foreseeable future.

At this stage, Adobe’s AI tool cannot produce long-form videos. It can only produce five seconds, and tops at 720p and 24 frames per second.

Its competition, OpenAI, has said that Sora, the company’s video generation tool, can generate videos up to a minute long, but it is not currently available to the public.

Adobe has also said its AI has been trained on material the company was permitted to use, as other generative AI tools have come under scrutiny for using discarded YouTube videos to train their systems.

Meta was also criticized because they have allegedly used users’ personal videos to train the generative AI tools.



Leave a Reply

Your email address will not be published. Required fields are marked *