Vectorview helps to evaluate the performance and security of language models. Targeted testing with real-world scenarios is supposed to detect and prevent unintended behavior that is often missed by generic benchmarks. Sources: TechCrunch, Y Combinator
Assembly AI introduces its new Universal-1 speech recognition model, which is said to have 30% fewer hallucinations in speech data and 90% fewer hallucinations in ambient noise compared to OpenAI’s Whisper. The model offers improved accuracy for English, Spanish, French, and German, supports code-switching, optimized timestamp estimation, and faster parallel processing, which can be beneficial …
A demo video on Twitter/X shows a function commonly known as “inpainting”: parts of the image can be selected with a brush tool and then changed by text command. Source: Axios
OpenAI presents its new AI technology “Voice Engine“, which can apparently imitate human voices deceptively realistically. However, the company is limiting access to selected partners for now. Source: VentureBeat
MyShell TTS introduces OpenVoice. The tool can replicate a person’s voice in multiple languages using short snippets of audio. OpenVoice allows detailed control over voice style, emotion, accent, rhythm, pauses, and intonation. Source: Hacker News
Resemble AI introduces Rapid Voice Cloning, a tool that creates AI-powered voice clones from short audio recordings in less than a minute. Source: VentureBeat
Adobe announces generative AI solutions to streamline the enterprise content supply chain. Adobe GenStudio provides marketers with an AI-powered offering to quickly plan, create, and manage brand-compliant content. Seamlessly integrated with the Adobe Firefly image generator, Adobe says it enables scalable content production with new Firefly services and custom models.
Opera now lets you download and run AI language models locally on your own computer – without an Internet connection, at no extra cost, and privately. There are over 150 models from more than 50 families to choose from, including Llama from Meta, Gemma from Google, and Vicuna. The feature is initially available to Opera …
Stability AI has released Stable Audio 2.0, an update to its generative audio AI. The new version can create audio clips of up to three minutes from text descriptions. Stable Audio 2.0 can also transform uploaded audio files based on natural language instructions. The company has seemingly taken copyright protection very seriously: It says it …
Merging, in the context of generative AI, refers to the combination or fusion of different AI models or their characteristics. Similar to creating a collage, the best or desired features of multiple models are united into a new model. A practical example is the merging of different Stable Diffusion models, where one model’s ability to …