What needs checking when a video model promises a 15-second clip in less than 15 seconds?
Pika says it has released Pika H3 Flex, an open-weight video model that generates 15-second clips with synced audio, faster than real time, at up to 4K resolution. It also lists exportable audio stems—separate audio tracks that can be taken into an editing workflow.
“Open-weight” means the model’s learned parameters are made available. For business use, checking the licence remains a separate step.
The practical hook is the combination: clip generation, sound aligned with the video, and audio that can be exported separately. For someone evaluating a production workflow, each deserves its own check.
What to watch next
Test generation time at the intended resolution, check whether sound stays aligned with the action, and try the exported stems in an editor. Treat “faster than real time” as Pika’s claim until tested under the conditions the workflow requires.
Before building a production process around it, confirm that the speed, audio exports and licence fit the intended use.