49 pointsby victormustar4 months ago

4 comments

Tiberium4 months ago

As for what it is: it is a multimodal LLM that can accept both text and images, and generate both text and images as output.

minimaxir4 months ago

Note: this model was just released from DeepSeek. <a href="https://github.com/deepseek-ai/Janus">https://github.com/deepseek-ai/Janus</a>

Later discussion: <a href="https://news.ycombinator.com/item?id=42843131">https://news.ycombinator.com/item?id=42843131</a>

pieix4 months ago

Anybody used this yet and can share example outputs?

Janus-Pro: Autoregressive framework unifying multimodal understanding&generation