When I use AI to generate drafts from videos, the process doesn't finish and stays in the generating state. What should I do?

If AI video-to-draft generation fails to complete, the "safety filter" built into the AI model may be activated. ## What is the safety filter? The safety filter is a restriction function designed to block content related to weapons (chemical weapons, explosives, etc.). Generation may fail due to false detection of the safety filter in videos containing the following terms or contexts: - Terms related to chemicals or pharmaceuticals - Explanations of procedures for toxicity tests or safety tests - Technical terminology related to biology or chemistry This filter is set according to standards established by the AI model provider, so Teachme Biz cannot individually disable it. We are considering continuous improvements to enhance accuracy. ## Solutions If generation fails due to the safety filter, please try the following three methods: *For troubles not related to the safety filter, please check:* [Common Troubleshooting for AI Video-to-Draft Creation](https://help.teachme.jp/hc/en-us/articles/34226455240857) ### Use "Split by Actions" mode When generating video manuals, select the "Split by Actions" mode. This reduces dependence on audio information, increasing the probability of successful generation by avoiding the filter. ### Make prompts (instructions) more specific By crafting the prompts you enter during video manual generation, you can help the AI correctly recognize legitimate business processes. Examples: "This is a cleaning procedure for [product] in a factory" (clearly indicating industrial use) "This is an official inspection process based on ISO standards" (indicating compliance with standards) ### Use videos that don't contain relevant terms If the audio contains words that are likely to trigger the safety filter, re-recording the video while avoiding these terms will increase the probability of successful generation. Please be assured that specialized terminology or detailed captions/explanations that might trigger the safety filter can be manually edited and added after draft generation.