NVIDIA's Audio Flamingo Next: Open audio-language models / hacker news