Problems with multi-scene video audio generation and audio length

[di-osc]

Hi,

Thank you for your great work on this project!

I’ve encountered a couple of issues while using the audio generation feature:

1. Multi-scene video audio generation
   When generating audio for videos containing multiple scenes, the results are often not ideal. In particular, when no prompts are provided, the generated audio sometimes includes strange or irrelevant sounds.

2. Audio output length
   When generating audio for a 20-second video, the output audio is only about 15 seconds long, and the remaining portion is silent.

Could you please let me know if these are known limitations or unexpected behaviors?

Thanks in advance for your help!

[GinChow]

Thank you for your feedback!

1. We recommend using single-scene videos, as frequent scene transitions can affect the semantic quality of the generated audio. Additionally, providing brief text prompts—just a few words like "gunshot," "footsteps," or "wind blowing"—will yield better results than leaving the text field empty.

2. Regarding output audio length, we've set a maximum limit to ensure optimal quality. For best results, we recommend keeping your input video between 5-10 seconds long.