Add audio to a silent video

Take a silent clip and give it sound. Three different jobs hide behind that ask: synchronized sound effects (footsteps, rain, impacts that line up with the action), spoken narration, or a music bed. Each routes to a different model, and only one of them is a dedicated SFX model. This skill covers the SFX route directly and hands the other two to their sibling skills.

Inputs to collect

• The silent video (URL or UUID). For SFX, max 10 seconds per call.
• What kind of audio: sound effects synced to the action, narration/voiceover, or background music. This decides the route.
• For SFX: an optional prompt describing the sound to bias toward (e.g. "heavy rain on metal, distant thunder"). Leave it off to let the model read the video and design sound from what it sees.
• Whether the result must be the video with sound muxed in (SFX route returns this) or a standalone audio file to mix yourself.

Models

• Sound effects → Mirelo SFX 1.5 (mirelo:1@1) - the one dedicated SFX model. Feed it a video and it returns the same video with synchronized sound effects muxed in. Honest caveat: dedicated SFX coverage is thin (this single model), so set expectations and consider the native-audio route below.
• Narration / spoken voiceover → use the voiceover skill (text-to-speech). Not this model.
• Music bed → use the music skill (text-to-music). Not this model.
• Often the better route: native audio at generation time. Several video models emit incidental sound (ambient, foley, even speech) as part of the original generation. If you still control the generation step, turning that on usually beats bolting SFX on afterward. Check capabilities and the audio flag via runware-models + runware-run before regenerating.

Confirm the live model and its schema via runware-models + runware-run before calling - never hardcode a stale choice.