Where Is This Wild Leopard Cub's Mom? | Dodo Kids
empathyacknowledged0:27–0:32
The cub cries for its mom, expressing distress.
Watch at 0:27 ↗Data for children's media
Kidoio codes what kids' videos model and reward, second by second, against established child-development frameworks. License the timestamped moments, whole-video scores and age bands to evaluate safety models, set brand-safety rules or run research.
Figures as of October 2026. Labels only: video IDs, timestamps and descriptions. No footage, no personal data.
Every moment links to the second it describes. Open one and watch what was coded.
The cub cries for its mom, expressing distress.
Watch at 0:27 ↗An advertisement for a tablet interrupts the educational content.
Watch at 2:06 ↗Adult characters grab snacks from the cart without waiting for the child, leaving her with nothing.
Watch at 0:47 ↗Nine developmental dimensions per video, a recommended age band, and timestamped moments with a description of what happens on screen.
| Dimension | What is coded | Framework | Moments |
|---|---|---|---|
| Concerning moments | Ads interrupting content, scary imagery, unsafe or mature material, with the youngest ages affected | AAP / WHO screen-time guidance | 3,682 |
| Empathy | A feeling or need noticed, acknowledged or rewarded | RULER (Brackett), Gottman | 3,474 |
| Aggression | Force or intimidation, and whether it works | Bandura social learning, Tremblay | 2,796 |
| Adult voice | How adults speak to children: authoritative, authoritarian, permissive | Baumrind parenting styles | 2,785 |
| Self-regulation | Strategies modelled for waiting, frustration or fear | Gottman emotion coaching | 1,672 |
| Body autonomy | Consent, privacy and personal space | AAP guidance | 154 |
| Cooperation, conflict resolution, reward framing, cognitive engagement | Scored for the whole video | Bandura, Vygotsky | — |
Licences cover the annotations, never the footage. Prices start where shown; every quote is sized to the cohort you need.
A cohort of videos and moments to test and benchmark kids-suitability or safety classifiers. 1 year, not exclusive.
Train production models on the labels. 1 year. Labels only, never footage.
Whole-video scores and timestamped moments for any YouTube ID; new videos analysed within 24 hours.
Lists of channels and videos that meet set rules, such as no rewarded aggression, no ad interruptions, ages 3–5.
Kidoio's analysis of your own catalogue, with a summary report.
Any cohort for research and publication. One lab, 2 years, cite Kidoio.
CSV or JSONL, one row per moment or per video score, with a methodology manifest and ML Commons Croissant metadata, ready for Hugging Face. Datasets are versioned and frozen at publication, so a licence always points to the same rows.
Videos are publicly available, mostly on YouTube, chosen by parents in Kidoio's apps and by Kidoio's team. A multimodal AI pipeline watches and listens to each video and codes every scene against Kidoio's methodology (kidoio-v1.0), developed with child-development researchers.
Each label carries a review status. The current corpus is model output; a two-reviewer gold set of 500 videos with published agreement figures is in preparation.
Kidoio licenses its annotations only: video IDs, timestamps, labels and descriptions. It does not distribute footage, frames or thumbnails, and the data contains no personal data of children or app users.
Real rows from the current corpus, in the delivery format.
Say which ages, dimensions and volume you need, and what the data is for. We reply with a cohort and a quote.