阿里巴巴 Wan 2.5 API — 支持音画同步的 AI 视频生成解决方案

When adding speech, don’t just request “dialogue.” Instead, provide the exact words to be spoken and specify who says them. This is especially important in multi-character scenes where order and clarity matter. For example: Character A: “We have to keep moving.” Character B: “Not until we find shelter.” By writing dialogue this way, you ensure the API assigns the right lines to the right characters.

In some videos, the atmosphere should be driven by visuals or sound effects alone. If you don’t want dialogue, make that clear in your prompt. Adding phrases such as “no dialogue” or “no actors speaking” prevents unintended voices from appearing. This small detail keeps your output aligned with the creative vision.

Beyond dialogue, ambient sound and music set the emotional tone. Be specific about the kind of environment or soundtrack you want, whether it’s natural or dramatic. Examples include: “soft rain tapping on windows with distant thunder” or “fast-paced action music with heavy percussion.” The clearer you are, the better the model can synchronize visuals with sound to create an immersive result.

Wan 2.5 excels when prompts include setting, lighting, camera perspective, and mood. Instead of writing “a person walking on a road,” expand the description to capture cinematic elements. For example: A wide shot of a mountain road at sunset, golden light flooding the sky, a cyclist racing downhill, with energetic background music in the background. This depth of description allows the API to produce more natural, dynamic, and visually coherent videos.