Advertisement

Pika Labs sound effects now available for all users — I tried it and here is how it sounds

 Pika Labs.
Pika Labs.

Pika Labs, the artificial intelligence video platform is making its sound effects feature available to all users. It allows creators to easily add sound to any video, either before or after generation.

First launched for Pro plan subscribers, the sound effects come from a Pika Labs in-house text-to-audio model that tries to match the video prompt text.

This takes AI video to a new dimension, expanding on the character voice and lip synching tools previously added to the Pika Labs platform.

Other players are also experimenting in the AI video with sound space. ElevenLabs has its own sound effects library and Runway has a limited release lip sync feature.

Putting Pika Labs sound effects to the test

There are two ways to add sound effects to a video within Pika Labs, as part of the initial prompt by toggling the sound effects switch, or after the fact with a separate prompt.

I’ve tried both for this hands on review and found that unless you need to pick the sound, such as specifying a time of year or specific noise, the first approach gives better results.

1. The bar at Christmas

This is based on an image I’d created for a previous story about the Suno AI music app, depicting a bar on the moon at Christmas. I left the text prompt blank and toggled the Sound Effects switch to on to see how well it would match the image.

It did a surprisingly good job. Somewhat loud but then so are some of the best bars. It didn’t really capture the Christmas element but the only indication was a tree in the corner.

2. The supercar

For the first test I gave Pika Labs the following prompt and enabled Sound Effects, allowing the model to follow both the prompt and the video: “A sleek, red sports car racing down a city street at night, leaving a motion blur trail behind it.”

It captured the movement well and did generate a sound effect of an engine revving, but it was surprisingly quiet. I had to ramp up the sound in iMovie to hear it properly but it sounded good overall.

3. Splashing in a pool

This is known as the splash test. How well it not only handles water sounds but how it does so when timed to a video. I used the prompt: “A person in mid-air, silhouetted against a clear blue sky, about to dive into a sparkling swimming pool.”

It didn’t follow the prompt particularly well in terms of the visuals, though it looked roughly along the lines of the prompt. It made the character more mermaid than person but the splash sound was good as were other general water noises.

4. The busy kitchen

I was really impressed with this image I generated in MidJourney. It almost felt completely out of focus, which was impressive for an AI model — but has just the right focus on the food, putting the chefs in the background.

This made it the perfect candidate to test Pika Labs sound effects. I simply gave it the image and selected sound effects then hit the generate button. It captured movement well and created sounds just like a working kitchen.

5. The fireworks display

The final test was one I've used on every sound effect test so far — fireworks. It should be easy but the challenge is syncing the sound to the explosion.

I used the prompt: "A vibrant firework explosion illuminating a night sky, with sparks in various colors radiating outward." It didn't do a bad job considering there was no manual syncing at all, it was entirely done by the AI model.

Conclusion

Sound effects add an important dimension to artificial intelligence video content, well in fact they add a significant dimension to any video content.

You can almost hear the sound of a car revving its engines just by looking at the picture and so its vital the sounds are not just correct but synced properly.

That is something Pika Labs seems to have achieved with their tool. It isn't perfect, doesn't always get exactly the noise you want but it is designed with enough flexibility that it can be used as a throwaway idea, or for a full, orchestrated production.

More from Tom's Guide