Frame-level scanning
Frames are sampled across the video and each one is classified independently as safe or NSFW.
CatalogLatency Async; scales with video duration and sampling rate Async true Rate Limit Per API key Max Input Scales with video duration
cn8/studio-nsfw
Give the service a video URL and it samples frames across the clip, classifies each
Learn more about NSFW Detection Videoasync (job)Instant previewvideoanalysismoderation
Try it
Video to scan
Scan density (frames / second)
Result
The result shows up here
Upload your video on the left and hit Run.
Overview
Give the service a video URL and it samples frames across the clip, classifies each one, and returns the timestamps where it finds NSFW content — along with the video's duration and how many frames were checked. You can tune how often frames are sampled with target_fps. This is an async job: submit the video, poll until it's ready, then read the result. Great for moderating uploads, enforcing content policies, and running pre-publish safety checks.
Key capabilities
Timestamp output
Returns the timestamps (in seconds) where NSFW content was found, plus a count and a has_nsfw flag.
Sampling stats
The result reports the video duration, the target_fps used, and how many frames were actually sampled.
Adjustable sampling rate
A higher target_fps catches more but takes longer; the default of 2.0 fps is a good balance for most videos.
When to use it
Upload moderation
Scan user-uploaded video before it goes live, and block or flag clips that return NSFW timestamps.
Content compliance
Make sure video meets your platform's safety policies before distribution.
Batch scanning
Run a library of videos through the service to surface the ones that need manual review.
Input & output
input
A video URL and an optional target_fps (default 2.0).
JSON body
output
NSFW timestamps, video duration, and sampling stats (async job result).
JSON
Guides & tips
How it works
- The service samples frames across the video at your target_fps and classifies each sampled frame as safe or NSFW.
- It collects the timestamp (in seconds) of every frame flagged as NSFW.
- The result returns those timestamps along with the total duration and how many frames were sampled — so you can decide how to act on them.
Choosing target_fps
- Default 2.0 fps: a good balance for most content — one frame every half second.
- Higher (e.g. 5.0): catches brief NSFW moments, but takes longer to process.
- Lower (e.g. 0.5): faster, but may miss short segments.
Specs
Schema
Request body
videoUrlrequiredstringURL of the video to scan.
target_fpsnumber · default: 2How many frames per second to sample (default: 2.0). Higher catches more but takes longer.
Response
video_meta.durationnumberTotal video duration in seconds.
sampling.target_fpsnumberThe sampling rate that was used.
sampling.sampled_framesintegerHow many frames were actually sampled.
nsfw.timestampsarrayArray of {start, end} objects (in seconds) where NSFW content was found.
nsfw.countintegerNumber of NSFW detections.
nsfw.has_nsfwbooleanTrue if any NSFW frame was found.
Pricing
Billed per video analyzed.
| Service | Unit | Price |
|---|---|---|
| NSFW Detection | item | 4 credits/video |
Priced per video, regardless of duration or sampling rate.