Create AI video with supported audio options

Some RedVideo models manage audio as part of the generation, some expose a user-controlled audio toggle, and others return video without generated sound. The model picker and active controls show the current behavior.

Wan 2.6 Flash currently supports an optional audio setting for image-to-video. Other supported Wan, Veo, Kling, Seedance, Core, and HappyHorse routes apply their own audio rules, which can also change by quality or workflow.

When a model supports generated audio, include only the sound information that matters to the shot: ambience, an action cue, a short spoken moment, or the absence of a distracting sound. Keep visual direction clear so audio detail does not replace the scene description.

Generated sound is probabilistic. Speech, synchronization, pronunciation, music, and environmental audio may not match the prompt exactly and should be checked before use.

• Name the visible source of an important sound
• Keep dialogue concise when the model supports it
• Review synchronization and unintended audio artifacts

An audio toggle asks a supported model to generate or include sound; it does not mean the workflow accepts an uploaded audio file. Reference audio is a separate input available only on workflows that explicitly expose reference-audio slots.

Current Seedance 2 reference-to-video routes can expose supported audio references. Select the reference workflow and review its duration, format, and total-length validation before uploading any source audio.

Audio can affect the generation quote or remain separately charged under certain access modes. For example, optional automatic audio can still use credits on otherwise eligible Unlimited Core runs.

The studio displays the active audio control, model access, and current quote. Export and review the completed video on the devices and platforms where the soundtrack will actually be heard.