What's this about?
Pika introduced Pika Audio: four frontier foundation models covering what it calls the full spectrum of generative sound, priced below every audio model on the market — up to 20 times cheaper — with a joke footnote that there is "literally no disclaimer" attached to that claim.
What exactly is "the full spectrum of generative sound"?
Audio generation is not one problem. Speech synthesis turns text into a voice. Music generation composes and performs. Sound-effect generation produces the door creak or the footstep. Ambient or foley generation fills a scene with its environment. These have historically been separate models with separate training data, which is why video tools often generate beautiful silent footage. Covering all four with one family is a bet that the next bottleneck in generative video is the soundtrack.
Why is this worth noticing?
- Pricing is the actual announcement. A 20x reduction moves audio generation from a per-project cost you think about to something you can run continuously.
- It fits the pattern of the last week — Ultrafast, Gemini 3.7 Flash, Grok 4.6 — where the competitive move is not higher capability but the same capability at a fraction of the price.
- Audio has lagged video and text in the generative wave, largely because the data is harder to source cleanly; a lab claiming full coverage suggests that constraint is easing.