[June 6 Big Model Daily Collection] News: ByteDance launches Seed-TTS text-to-speech model, which can generate human-like speech; collaboration platform Asana launches “Al Teammates” feature to help teams arrange and plan tasks;

ByteDance launches Seed-TTS text-to-speech model, which can generate human-like speech

Link: https://news.miracleplus.com/share_link/29185

ByteDance’s team recently proposed a technology called Seed-TTS, which is a series of large-scale autoregressive text-to-speech (TTS) models that can generate speech that is almost indistinguishable from human speech. Seed-TTS is a basic model for speech generation, which performs well in speech context learning and matches real human speech in objective and subjective evaluations in terms of speaker similarity and naturalness.

Collaboration platform Asana launches “Al Teammates” feature to help teams arrange and plan tasks

Link: https://news.miracleplus.com/share_link/29186

Office collaboration platform Asana said in a press release that its artificial intelligence model can use the stored team history and past project information to assign work to people with the best skills, such as marking designers who understand the brand style to work on creative projects. The feature called “Al Teammates” can help teams arrange tasks and determine what responsibilities each person has, make suggestions and plans.

Kuaishou version of Sora “Ke Ling” open test: Generate more than 120s video, understand physics better, and accurately model complex movements

Link: https://news.miracleplus.com/share_link/29187

A new domestic video generation model called “Ke Ling” was recently released. It uses a similar technical route to Sora, combined with a number of self-developed technical innovations, the generated video not only has a large and reasonable range of motion, but also can simulate the characteristics of the physical world, with strong concept combination capabilities and imagination. From the data, Keling supports generating 30fps ultra-long videos up to 2 minutes long, with a resolution of up to 1080p and supports multiple aspect ratios. Keling is not a demo or video result demonstration released by the laboratory, but a product-level application launched by Kuaishou, a leading player in the short video field. Moreover, it focuses on being pragmatic, not issuing empty promises, and launching online as soon as it is released. The large model of Keling has officially opened invitations for testing in the Kuaiying APP.

The boss of Stable Diffusion ran away to open a new pit, and the abandoned SD3 open source became a mess

Link: https://news.miracleplus.com/share_link/29188

The new company of the former CEO of Stability AI has finally surfaced: Schelling AI, as he said when he resigned and ran away, is committed to “decentralized artificial intelligence system”. This news was announced by Emad Mostaque himself in his latest public appearance. The official website domain name schelling.ai will currently jump to the official 𝕏, which is still an empty number. The only news at the moment is that the new business is related to large language models, and the business direction will focus on government affairs, healthcare and education, in other words, toB.