When I use AI to generate a draft from a video, the process stays in a generating state and never completes. What should I do?

2026-02-19 11:04:57 UTC 2026-03-23 08:09:46 UTC

If draft generation from a video using AI does not complete, the "safety filter" built into the AI model may have been triggered.

What is the safety filter?

The safety filter is a restriction feature designed to block content related to weapons (such as chemical weapons and explosives).
Generation may fail due to false positives from the safety filter in videos that contain terms or contexts such as the following.

  • Terms related to chemical substances or reagents
  • Descriptions of procedures related to toxicity testing or safety testing
  • Specialized terminology related to biology or chemistry

Because this filter is based on standards set by the AI model provider, Teachme Biz is unable to disable it on an individual basis.
We are continuously exploring improvements to increase accuracy.

Workarounds

If the safety filter is triggered and generation does not complete, please try the following three methods.

- For issues not related to the safety filter, please refer to the article below.
Common issues with AI draft creation from videos

Use "Split by Action" mode

When generating a video manual, please select "Split by Action" mode.
This reduces reliance on audio information, increasing the likelihood of successfully bypassing the filter and completing generation.

Make your prompt (instruction text) more specific

By refining the prompt you enter when generating a video manual, you can help the AI correctly recognize the content as a legitimate business process.

Example:
"This is the cleaning procedure for ○○ in a factory." (Explicitly stating the industrial purpose)
"This is a formal inspection process based on ISO standards." (Explicitly stating the applicable standard)

Use a video that does not contain the relevant terms

If the audio contains words that are likely to trigger the safety filter, re-recording while avoiding those words can increase the likelihood of successful generation.
Please note that specialized terms subject to the safety filter, as well as detailed captions and descriptions, can be manually edited and added after the draft is generated.

Asking the chatbot a question