AI moderation

Content moderation workflows

Build content moderation workflows with FastPix: pre-publish safety gates, human review and appeals, restricted playback, and fixing flagged media with transforms.

Content moderation is most effective when it is part of a publishing workflow, not a one-time check. FastPix analyzes your video or audio, returns the categories, confidence scores, and time-stamped segments it detects, and notifies your application through the video.media.ai.moderation.ready webhook when the analysis is complete. Your application then decides what happens to the media.

You can integrate any of the workflows below into your own app, product, or publishing pipeline, and adapt them to your audience and content policy. Each one shows a pattern you can build with FastPix moderation results, followed by how to use FastPix APIs to fix flagged media.

The building blocks are the moderation result (a verdict, per-category scores, and time-stamped segments), the video.media.ai.moderation.ready webhook, and the media accessPolicy you update to control playback. For what the scores mean and how to pick thresholds, see Content moderation.


How a moderation workflow works

Most content moderation workflows follow the same sequence:

  1. A video or audio asset is uploaded or selected for analysis.
  2. Moderation runs asynchronously while the asset stays in a pending or restricted state.
  3. Your application receives the completed results on the video.media.ai.moderation.ready webhook.
  4. Your rules evaluate the detected categories and scores.
  5. The application allows the media, sends it for review, restricts access, or edits it to fix a flagged moment.

Treat moderation results as decision support. A simple, durable model is to map score bands to three actions:

  • Allow when all findings are below your review threshold.
  • Review when findings fall in an uncertain range that needs human judgment.
  • Restrict when a category crosses your high-risk threshold.

Keep these decisions separate from the moderation result itself. The result records what was detected, and your policy decides the action, so you can change thresholds later without reworking your integration.


Pre-publish safety gate

Use a safety gate when media must be reviewed before it becomes available to viewers.

  1. Keep newly uploaded media unavailable or restricted.
  2. Start moderation after the media is ready.
  3. Wait for the completed moderation result.
  4. Publish only when the result meets your policy.
  5. Send uncertain or high-risk results to review instead of publishing automatically.

This workflow suits OTT and streaming services, online learning and kids’ platforms, and enterprise or internal media libraries, where content is reviewed before it becomes available to viewers.


Human review and appeals

Use human review for borderline content, appeals, or situations where context matters.

When a result enters your review range, create a review task that contains:

  • The media and the relevant time range.
  • The detected categories.
  • The confidence scores.
  • The reviewer decision options.
  • A reason or note field for the final decision.

Include a short-lived signed playback URL so a reviewer can watch the media without making it public. A reviewer can approve the media, restrict it, request an edit, or escalate it for a second opinion.

If a creator appeals a decision, send the same review record to a reviewer along with the original finding, the policy used, and any supporting context. Human review is especially valuable for satire, news, documentary footage, education, fictional scenes, and content that may be acceptable under your community guidelines.


Restricted playback workflow

Not every finding requires deleting the media. For medium-confidence findings, keep the asset stored while limiting who can play it by updating the media accessPolicy and requiring signed playback or DRM.

This workflow suits:

  • Age-gated content.
  • Content awaiting review.
  • Content visible only to the uploader.
  • Content available only to trusted users.
  • Content under appeal.

After a reviewer approves the media, restore normal access. If the reviewer rejects it, keep it restricted or remove it according to your policy. Restricting access rather than deleting preserves the option to appeal or review.


Fix flagged content with FastPix APIs

When your policy calls for an edit, FastPix can create a corrected media version from the moderation result. The general pattern is: identify the relevant category and time range, apply the matching media transformation, then run moderation again before publication. The same pattern works with segment removal, clipping, audio replacement, audio overlays, and stitching.

Run moderation again on every corrected or newly combined version before you publish it, and treat that version as a new moderation decision so your audit trail stays clear.

Remove a flagged moment

Use this workflow when a short section of violence, unsafe imagery, or another policy violation should not appear in the published version.

  1. Receive the moderation finding and note its start and end time.
  2. Keep the original media restricted while the edit is prepared.
  3. Use Remove unwanted video segments to remove the flagged interval while preserving the rest of the video.
  4. Use the resulting media asset as the candidate for publication.
  5. Run moderation again before making the corrected version available.

This is useful when the surrounding content is valuable and only a specific moment needs to be removed.

Publish only a safe section

Use Clip and trim videos when you want to publish a short, approved portion rather than the full source media. For example, create a highlight from a reviewed range, publish the new clip, and leave the original media restricted until a broader decision is made. A clip is a separate media asset, so it can have its own playback access and moderation status.

Replace or cover problematic audio

Use this workflow when the visuals are acceptable but the spoken audio contains profanity, a harmful statement, or another audio finding.

  1. Identify the affected portion from the audio moderation result or transcript.
  2. Prepare a clean replacement, such as edited speech, silence, or approved background audio.
  3. Use Replace a video’s audio track when the original audio should be replaced, or Overlay audio on a video timeline when audio should be placed over a specific part of the timeline.
  4. Run moderation again on the corrected media before publication.

This lets you preserve the visual story while correcting an audio issue.

Create a clean compilation

If a reviewer approves several safe sections, create separate clips for those ranges and combine them with Merge and stitch videos. Publish the result as a new, reviewed version while the source remains available only to authorized reviewers.


Combine moderation with other video insights

Reviewers often need more context than a category and a score. Pair moderation with other In-Video AI insights, such as:

These insights help reviewers decide faster, but they should support the moderation result rather than override your content policy.


Built for content moderation workflows

FastPix gives you everything these workflows need in one video platform. AI moderation analyzes both video and audio and returns a verdict, per-category scores, and time-stamped segments. The video.media.ai.moderation.ready webhook and moderation result drive your decisions, while accessPolicy and signed or DRM playback control who can watch. A full set of media transforms, including removing segments, clipping, trimming, replacing or overlaying audio, and merging and stitching, lets you fix flagged content and republish it.

Because moderation and media processing run on the same platform, you can go from detection to a corrected, re-checked version without stitching together separate tools. Your application stays in control of the policy: publishing rules, review queues, access changes, notifications, and audit storage.


Frequently asked questions

What is a content moderation workflow?

A content moderation workflow runs moderation on a media asset, evaluates the returned categories and scores against your policy, and takes an action such as publishing, restricting, routing for human review, or fixing the flagged media. With FastPix, the video.media.ai.moderation.ready webhook triggers your rules once the analysis completes.

How do I gate uploads before they publish?

Keep newly uploaded media restricted, run moderation once the media is ready, and publish only after the result meets your policy. Route uncertain or high-risk results to human review instead of publishing automatically.

Can I fix flagged media instead of deleting it?

Yes. Use the segment start and end times to remove the flagged interval, clip an approved section, replace or overlay audio, or stitch safe sections together, then run moderation again before publishing. See Fix flagged content with FastPix APIs.

How should I handle appeals?

Keep the media restricted rather than deleted, and send the original finding, the policy used, and any supporting context to a reviewer. Restricting access preserves the option to appeal or restore.


What’s next

Turn a finding into a fix instead of just blocking the asset: