Google launched the open-source Gemma 4 12B AI model, designed to run multimodal tasks such as text, images, and audio locally on ordinary laptops
ElevenLabs has launched a beta version of its AI-powered Dubbing v2 model, which aims to preserve the original emotion and performance of translated speech across over 90 languages. This new model utilizes direct voice adaptation instead of text and incorporates synchronization-aware translation for better alignment, and is currently available through ElevenCreative and ElevenProductions. Google has also released Gemma 4 12B, an open-source AI model capable of running multimodal tasks like text, image, and audio processing locally on standard laptops with 16GB of RAM. This model offers similar performance to its larger 26B counterpart while being half the size and is the first medium-sized model to support native audio processing, available for commercial use under an Apache 2.0 license. AI
IMPACT New multimodal and translation models offer enhanced local processing and cross-lingual capabilities, potentially lowering barriers for content creation and AI application development.