dsh-multimodal

About

Give DeepSeek a pair of eyes and a paintbrush: paste screenshots/images directly in the session, the GLM vision model first accurately transcribes the image content (error messages, code, interface preserved word-for-word), then DeepSeek continues to process your question - completed in the same turn, completely seamless; when illustrations are needed, DeepSeek automatically calls the text-to-image backend to generate images and display them in the session.

Description machine translated from the original.

Topics