Image Generation

Today's Stories

1. 1

   Google's Lyria 3.5 lets users edit song sections without restarting

   Google released Lyria 3.5, a music generation model that produces more natural-sounding melodies, better lyrics, and more realistic vocals. The model is now available through Google Flow Music and includes a new feature called 'Selective Section Painting' that allows users to edit specific parts of a track or turn short melodies into full songs without starting over. The ability to edit individual sections without restarting the generation process makes music creation faster and more flexible for users. Users can also fine-tune tempo and duration of specific elements like vocals, drums, and bass, giving them more creative control over the final output.

   Tracks can run from 30 seconds to 3 minutes in length. Google has not provided details about the training data for Lyria 3.5, despite having disclosed that Lyria 3 was trained on materials YouTube and Google had the right to use under their terms of service, partner agreements, and applicable law.

2. 2

   Google's SynthID watermark survives edits, but won't stop AI fakes

   Google announced that its SynthID watermarking technology, now used by OpenAI, Runway, and Nvidia, can survive aggressive image compression, resizing, and cropping—remaining detectable even after 300 rounds of simulated sharing. However, testing shows that a 20 percent crop is enough to break the watermark entirely, and the company acknowledges the technology is not designed to withstand deliberate adversarial attacks. AI content is being created at an unprecedented scale—Google's tools alone produced more than 100 billion AI images and videos in just a couple of years, far outpacing authentic content. While watermarks offer a way to label AI-generated material, they cannot solve the core problem: countless open-source AI models allow anyone to generate unlabeled content on their own computers, meaning watermarks will never catch all AI fakes. Access to SynthID verification is also limited to approximately 10 image checks per day, and different companies' watermarks don't recognize each other, making verification fragmented and unreliable.

   Google is exploring ways to make watermark verification more unified across platforms, but experts suggest the real focus should shift to authenticating truthful content rather than trying to detect all fakes. The company notes that SynthID's watermark is not invulnerable and could face a cat-and-mouse game of attacks and updates if it becomes a larger target.

3. 3

   China beauty firms use AI to slash product development from years to months

   Yatsen Holding, a Chinese beauty company, used artificial intelligence to compress a product development cycle from more than two years into six months for a new PDRN serum launched around China's Singles' Day shopping festival last fall. AI helped researchers narrow more than 50 ingredient candidates to three in a single week. Chinese beauty companies are using AI across the entire product pipeline—analyzing millions of social media posts to spot trends, identifying new cosmetic ingredients through biotech partnerships, and accelerating time-to-market. This speed advantage helps local firms close an innovation gap and respond faster to consumer demand.

   The pace of product launches from Chinese beauty companies may increase as AI-driven workflows become standard. The ability to reduce development cycles from years to months could reshape how quickly new formulations reach shelves across the region.

4. 4

   Hugging Face hosting thousands of deepfake nude tools

   Researchers at AI Forensics tested image editing tools on Hugging Face and found that 7 of 9 top tools easily converted clothed images of women into topless ones. In a separate test using a honeypot, they collected over 1,000 prompts in a week; 73 percent were sexual in nature, with 83 percent of those seeking to undress or sexualize submitted photos—95 percent targeting women, and 6.7 percent targeting apparent children. Hugging Face, valued in the billions, hosts around 5,000 AI image models capable of creating nonconsensual intimate images of real people, yet appears to lack platform-level safeguards. Unlike mainstream models from OpenAI and Google that use safety mechanisms called guardrails, the open-source models on Hugging Face can be easily misused with simple prompts—and are actively being used that way by thousands of people weekly.

   Hugging Face did not respond to WIRED's questions about its content moderation. Some pages promoting nudifying services were removed after WIRED contacted the company, though the connection is unclear. The EU and UK plan to ban "nudify" apps by the end of the year.

5. 5

   Jack Conte on AI's role in the future of online creativity

   Jack Conte, the founder of Patreon, shared views on how artificial intelligence will shape the future of creative work on the internet in a video presentation. As the leader of a platform that helps millions of creators earn money directly from their audiences, Conte's perspective on AI's impact on creativity and creator economics carries weight for anyone building or relying on creative work online.

   The video is available on YouTube at the URL provided, offering Conte's full argument on how creators and the internet ecosystem may adapt as AI becomes more prevalent.

6. 6

   Midjourney acquires astrology app Co-Star, appoints founder as chief design officer

   Midjourney, an AI image-generation startup, acquired the personalized astrology app Co-Star in a deal that closed in spring. Co-Star's founder Banu Guler will remain in charge of the app and will also become Midjourney's chief design officer. Co-Star combines human insight, NASA data, and AI to deliver personalized daily horoscopes and compatibility checks. The acquisition signals Midjourney's expansion beyond image generation into consumer apps, and Guler's design expertise may help shape how Midjourney builds its first dedicated applications.

   Midjourney plans to develop its first apps, including one dedicated to image generation. Currently, Midjourney's AI models operate through the web and Discord; these new apps represent a shift toward standalone products. The deal terms have not been disclosed.