🪄Adobe's AI Learns New Tricks

Adobe unveils new AI video tools

Key Points 

• Adobe's new AI tools can generate video clips from still images and text.

• The Firefly model allows fine-tuning with "camera controls" and generates video from reference images, ideal for B-roll footage.

• Videos created with Firefly are limited to five seconds but offer "commercially safe" content, reducing copyright concerns.

👩🏻‍💻News - Adobe has previewed its upcoming generative AI video tools, including a feature that creates video clips from still images. This builds on the Firefly video model, showcased in April, which will enhance AI video and audio editing in Adobe's Creative Cloud applications. The latest teaser also highlights footage generated by Firefly’s text-to-video capabilities, announced earlier this year.

✨What's new and different?

• The tool enables users to create video clips from text descriptions and fine-tune the outcome with "camera controls" that mimic real-world camera angles, movement, and shooting distance.

• Adobe also showcased an image-to-video feature within the Firefly video model, which generates video clips based on reference images. This functionality is ideal for creating extra B-roll footage or filling in gaps during production.

• The quality of the generated videos is comparable to OpenAI’s Sora model, which Adobe is considering integrating as a third-party option within its Premiere Pro software. However, the videos created using the text-to-video and image-to-video features are currently limited to a maximum length of five seconds.

• One potential advantage of Adobe’s Firefly model over Sora is its assurance of being "commercially safe." Firefly is trained on openly licensed, public domain, and Adobe Stock content, which helps alleviate concerns about copyright infringement.

💁🏻‍♀️Where and when you can try it - The text-to-video and image-to-video features will first be accessible in beta through a standalone Firefly application later this year. Eventually, Adobe plans to integrate the new Firefly video model into its Creative Cloud, Experience Cloud, and Adobe Express applications.

SAG-AFTRA and women’s groups push for approval of AI safety bill SB 1047

Key Points 

• Women's groups NOW and Fund Her compare unregulated AI to early social media issues, while SAG-AFTRA highlights dangers such as deepfake pornography.

• SB 1047’s potential to set a strong precedent is fueling both support from advocacy groups and opposition from major tech companies.

♨️News - As California Governor Gavin Newsom considers whether to sign or veto the highly debated AI safety bill SB 1047, SAG-AFTRA and two women’s organizations are urging him to approve it. The performers union, the National Organization for Women (NOW), and Fund Her have all sent letters to Newsom, expressing concerns about leaving AI technology unregulated.

The letters from NOW and Fund Her compare unregulated AI to the early issues seen with social media. Meanwhile, SAG-AFTRA’s letter highlights the specific dangers AI poses to its members, such as nonconsensual deepfake pornography and misuse of performers' likenesses. 

These recent endorsements for the bill join support from organizations such as SEIU and the Latino Community Foundation. In contrast, former House Speaker Nancy Pelosi, Congressman Ro Khanna, venture capitalist Marc Andreessen, and AI expert Fei-Fei Li oppose the bill. Additionally, tech giants Google, Meta, and OpenAI have also voiced their opposition.

🤔Why is it such a big deal? The US currently relies almost entirely on self-regulation and voluntary commitments to manage AI, and the industry prefers to keep it that way. As the first US AI safety regulation with real enforcement power, SB 1047 would set a strong precedent, which is likely driving both the support reflected in these letters and the intense industry opposition.