同时支持文本生成、图像生成、音频生成嘛？

[wangjue-wzq]

比如，输入文本、参考图、参考音频，让重新生成一段文本、图像、音频。

[Kira0096]

While the model currently supports text, image, and audio generation individually, we didn't chain these tasks together into single training samples in our dataset. That said, theoretically, this cross-modal capability can be unlocked through SFT.