Pika Soundtrack: video in, sound out. What Sume has instead

Pika Soundtrack takes an existing video and generates sound effects, voice, music and ambience for it. The Sume docs read 2026-10-01 describe no endpoint that does this. What they do have is a generate_audio flag on video creation and a Timeline soundtrack bed you lay under a finished cut.

What does Pika Soundtrack do?

Per Pika's August 18 post, you upload a video and it generates motion-aware sound effects, voice, music and ambience that follow the action, with an optional instruction about what to emphasize, include or leave out. It says the model is live in the Pika API Club. Pika's speed and benchmark statements are its own and are not repeated here.

What does Sume offer for sound on video?

Two separate things, neither of which reads a finished video and invents sound for it. See Video generation and Timeline 1.0.

How do I score a silent clip on Sume?

Make the sound as its own file, then place it. Generate music or voice with the audio endpoints, import it as Sume-hosted media, and give it to Timeline as the audio.url spine or the soundtrack bed. Timeline renders at $0.10 per output minute. For sound effects specifically, read sound effects for video without a SFX API.

Will the sound follow the action?

Not automatically. Timeline places the bed you supply; it does not analyze the picture to time impacts. Matching sound to motion is Pika Soundtrack's stated purpose, and the Sume docs make no equivalent claim. If sync matters, generate with a model that has audio on, or time your own audio against the cut.