The Multimodal Model Training team at Kakao has developed and released the Kanana-o family of Omnimodal Large Language Models (OLLMs), a suite of pretrained Text/Audio/Image-to-Text/Audio (TAI2TA) models, which is now available as an API. For more details on the Kanana-o family and its architecture, please refer to our previous technical blog posts. — 📝 이미지와 음성을 아우르는 카카오의 멀티모달 언어모델 Kanana-o 알아보기 — 📝 더욱 똑똑하게 답하며, 더욱 풍부한 감정표현을 향한 Kanana-o의 진화 과정 Kanana-1.5-o-9.8B-instruct-2602 is intended for research and application development in omnimodal understanding and text/speech generation tasks. — Image captioning — Document understanding — OCR-based reasoning — Automatic Speech Recognition — Text-to-speech — Speech translation — Multiturn conversation — Podcast-style speech generation — Omnimodal instruction following The model is optimized for both general-purpose and Korea-specific benchmarks, making it suitable for bilingual environments. All reported scores are computed after normalizing the original benchmark results to a 0–100 scale for cross-task comparability. > N/A — Prompt: indicates that the model did not provide the required prompt configuration for the corresponding…
Модальности:
Генерация текста Синтез речи Аудио Компьютерное зрение Мультимодальность
Области применения:
Следование инструкциям
Задача: Генерация текста
Автор: kakaocorp
Теги: omnimodal, multimodal, vision, audio, text-to-speech, speech-recognition, image-captioning, korean
Лайков: 4 | Загрузок: 0
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
