Describe a track. Hear it back.

Generated video usually arrives silent, which means a second tool, a second subscription, and a round trip every time the edit changes. Unify includes Suno V5.5 for full songs and ElevenLabs for music and voice in the same workspace as the video and image models, on the same credit balance. Generate the shot, generate the score, and put them together without exporting to a third service. For anything with narration — an explainer, an ad, a product demo — you can produce the voiceover next to the picture instead of booking it separately.

Music prompts work like image prompts: specificity is everything. Name the genre, then the instrumentation, then the tempo in BPM, then the mood, and finally the structural requirement. "Happy music" returns the model's average happy music. "Uplifting modern pop instrumental, bright plucked synth, four-on-the-floor kick entering at eight seconds, builds to a stop at thirty seconds for a voiceover tag" returns something you can actually cut to. If the track has to sit under dialogue, say so — asking for a sparse arrangement with space in the mid-range gets you a bed that does not fight the voice.

The practical workflow is picture first, then sound. Lock your shots, note the timings you need to hit, then describe the track around them: where it should build, where it should drop out, how long it runs. Because Veo 3.1 already generates native audio with its footage, decide early whether a shot carries its own diegetic sound or whether you are scoring the whole sequence — layering a generated score over generated ambience can get muddy. For podcasts and ads, generate the voiceover first and score around the read.