Want to know what sounds are in a video? The What’s That Sound? card in CNAPS Studio listens to the soundtrack and gives you labeled sound clips, a short scene summary, and a transcript with timestamps.
We made this card with our partner Cochl, whose sound AI does the detection. To run it, you need to sign in and connect your own Cochl API key.
From one video to sound clips and a readable summary
In the example shown here, the video features footage from the Perseverance Mars landing. The results include a clip labeled “Speech,” a scene summary, and a timed transcript of the landing announcement.
The transcript places the announcement at 00:04 and includes “Touchdown confirmed” and “Perseverance safely on the surface of Mars.” The summary describes the landing confirmation and lists keywords such as Mars, Perseverance, and touchdown.
These outputs offer different ways to review the same video: listen to a labeled clip, read a short overview, or use the transcript’s timestamps to locate speech. The labels and text are AI-generated, so compare them with the original recording before relying on a detail.
What you get
- Labeled sound clips. Review clips associated with the sounds detected in the soundtrack. The number of clips depends on the audio; a busy recording can produce more clips.
- A scene summary. Read a short description of the scene and what can be heard, together with keywords.
- A timed transcript. See the detected speech as text with timestamps and speaker labels, as shown in the sample.
You do not need to build a flow or connect individual models to use the prepared card.
How to try it
- Sign in to CNAPS Studio. Open the What’s That Sound? card.
- Connect your Cochl API key. Use the card’s Connect Cochl link for Cochl’s setup guide, then configure your key in CNAPS Studio’s API Key section.
- Upload a video. Add an MP4, MOV, or WebM file in the Input field. The card accepts videos up to 5 minutes long and recommends clear sound.
- Run and review. Click Run. The card estimates about 3–5 minutes. Play the sound clips and read the summary and transcript alongside the original video.
- Download your results. Stay signed in to download the results you want to keep.
Choose a video with sounds worth exploring
Try a pet video, an event recording, or a street scene. Start with a short clip whose sounds you can recognize yourself, so you can compare the labels with what you hear.
Clear sound gives the analysis a better starting point. If speech overlaps with music or other noise, review the transcript carefully. Treat speaker labels as a way to follow the transcript, rather than proof of someone’s identity.
If you only need a transcript for a batch of clips, see Speech Recognition (Clips).
Meet our partner Cochl
Cochl provides the sound detection used by this card. The partner panel describes support for more than 100 real-world sounds, including dog barks and breaking glass. CNAPS Studio brings sound analysis, a scene summary, and a timed transcript together in one prepared task.
Explore Cochl Sense or follow the Cochl setup guide to get started with your own key.
What sounds can it detect?
Cochl can pick out more than 100 everyday sounds, like dog barks, breaking glass, and sirens, along with speech. You can see the full list on Cochl Sense.
Do I need an account and an API key?
Yes. You need to sign in to CNAPS Studio and connect your own Cochl API key to run this card. Downloads also require sign-in. Cochl API usage follows Cochl’s pricing and usage policies.
Does this card add sound effects to my video?
This card analyzes sounds already present in your video. Its outputs help you review what was detected and what was said.
How long can my video be?
Up to 5 minutes. The card takes MP4, MOV, or WebM files, and a run usually takes about 3 to 5 minutes.
Upload a video and explore its sounds, summary, and transcript in one place.
