Summary
50+ hours of real infant cry recordings for training cry detection, sound event detection, and baby monitoring models. Every file is manually verified for clear cry audibility, with suspicious files that resemble internet downloads removed from the curated set. Captured in natural home conditions on smartphones, laptops, and tablets, with per-file metadata on location, background noise, and recording device
Introduction
This dataset contains 50+ hours of real infant cry audio, recorded in natural domestic settings: quiet bedrooms, noisy living rooms, both indoor and outdoor environments. Every recording was manually reviewed by an annotator to confirm that the cry is clearly audible and that the file is genuinely original, suspicious recordings that resemble material downloaded from the internet were excluded
Dataset Features
Scale & Quality
- 50+ hours of infant cry audio
- Manually verified – every recording reviewed for clear cry audibility
- Authenticity filter – files resembling internet downloads removed from the curated set
- Real domestic recordings – no synthetic audio, no augmentation, no AI-generated content
Audio Specifications
- WAV (majority) and M4A formats
- Sample rate: 48 kHz primarily, with 44.1 kHz and 16 kHz subsets
- Mono and stereo recordings
- File duration: 10 to 100 seconds per recording
- Recorded primarily on smartphones, with additional laptop and tablet captures
Metadata for Every File
- Recording location: indoor / outdoor
- Background noise level: quiet / moderate / noisy
- Recording device type: smartphone / laptop / tablet / external microphone
- Validation status: each file confirmed as a clear, original infant cry recording
Use cases and applications
- Infant cry detection in smart baby monitors, nursery cameras, and IoT devices
- Sound event detection (SED) models that include infant cry as a target class
- Sleep tracking apps that need to recognize cry events during the night
- Smart home assistants with cry-aware automation (notifying parents, lighting, audio response)
- Parental support apps that distinguish cry from other infant sounds (cooing, babbling, fussing)
Why this dataset solves real production challenges
- Significantly larger than academic alternatives. The most widely cited public infant cry dataset, CryCeleb, contains 6.5 hours of cry audio under a research-only license. At 50+ hours of curated recordings, our dataset removes the size and licensing barriers that block commercial deployment
- Manually verified, not scraped. A core problem identified in published literature is that existing open datasets contain noisy recordings and uneven distribution. Every file in this dataset was reviewed by an annotator, and any file showing signs of internet origin was removed
- Smartphone-first recording. Captured on the same class of devices that real baby monitor apps use: phones, tablets, and home microphones, not on studio equipment that does not match deployment conditions
Cleared for commercial use. Unlike the major academic infant cry corpora, which are restricted to non-commercial or research-only use, this dataset is licensed for commercial training of production ML systems
Sample dataset
A sample version of this dataset is available on Kaggle and HuggingFace. Leave a request in the form below for additional samples or the full version
Contact us
Tell us about yourself, and get access to free samples of the dataset
Didn't find what you were looking for?
Our collection includes many datasets for various requests
iBeta Level 1 Dataset
– 35,000+ videos
– 85+ participants
– zoom in and
zoom out
iBeta Level 2 Dataset
– 25 000+ videos
– 3D masks
– iBeta Level 2
iBeta Level 3 Dataset
– 10,000+ videos
– 12 Unique Masks
– iBeta Level 3
Display Replay Dataset for Liveness Detection
– 9,000+ videos
– 6,500+ participants
– Balanced mix of genders and ethnicities
