Extract TikTok reaction hooks with RenderIO scene detection

Works on both n8n Cloud and self-hosted instances. This template uses the n8n-nodes-renderio community node, which is installable on n8n Cloud and self-hosted setups.

Who's it for

UGC creators, performance marketers, and AI avatar producers who want to harvest the reaction shot from a TikTok and reuse it as the opening hook of new content. The extracted clip is designed to feed straight into Kling motion control or any AI avatar pipeline to mass-produce on-brand UGC reactions, without manually scrubbing through videos to find the cut point.

What it does

You paste a TikTok URL into a form. The workflow downloads the video, runs FFmpeg scene detection through RenderIO to find where the opening reaction shot ends and the product or app demo begins, then trims the source at that exact cut point. You get back a clean MP4 of just the reaction hook (typically the first 1 to 6 seconds), ready to drop into your avatar workflow. All video processing runs on cloud-based FFmpeg via RenderIO, so no local rendering is needed.

How it works

1. On TikTok URL Submission is a form trigger that accepts a TikTok URL from the user.
2. Download & Detect Scenes uses RenderIO to download the video and run an FFmpeg scene detection pass (select='gt(scene,0.3)') that writes timestamps of every detected cut to a scenes.txt file.
3. Wait 5 Seconds for Detection pauses to let the RenderIO job complete, then Verify Scene Detection Status checks the job state. If Scenes Detected branches based on whether the result is ready, with Wait 10 Seconds for Detect Retry handling retries.
4. Extract Initial Scene Cut fetches the scenes.txt file, parses all pts_time timestamps with a regex, and picks the first cut that falls in the 1 to 6 second window. If the first cut is too early it uses the next valid one; if no cut is found it falls back to 6 seconds.
5. Execute Visual Hook Split sends a POST to the RenderIO API with an FFmpeg command that trims the source from 0 to the detected hook end time, re-encoding with libx264 veryfast and AAC audio.
6. Wait 5 Seconds for Split and Verify Split Execution Status poll until the trim job finishes, with Wait 10 Seconds for Split Retry handling retries.
7. Show Hook Download Link displays a completion form with the hook end time, the detection reason, all detected cut timestamps, and a direct download link to the trimmed MP4.

Requirements

• An n8n Cloud or self-hosted instance
• The n8n-nodes-renderio community node installed via Settings > Community Nodes
• A free RenderIO account and API key from renderio.dev
• A public TikTok URL where the first 1 to 6 seconds is a reaction shot and the rest is the demo or payoff

How to set up

1. Install the n8n-nodes-renderio community node from Settings > Community Nodes.
2. Create a RenderIO API credential in n8n using your key from renderio.dev.
3. Import the workflow.
4. Open the On TikTok URL Submission node and copy the production form URL, or use the test URL for trial runs.
5. Activate the workflow.
6. Open the form, paste a TikTok URL where the first 1 to 6 seconds is a clear reaction (smile, gasp, cry, eyes wide) and the rest is product or app footage, then submit.

How to customize the workflow

• Change the scene sensitivity by editing the gt(scene,0.3) threshold in the Download & Detect Scenes FFmpeg command. Lower values like 0.2 catch softer cuts, higher values like 0.4 only catch hard cuts.
• Adjust the hook window by editing minS and maxS in the Extract Initial Scene Cut code node (defaults are 1 and 6 seconds).
• Tweak the output encoding in the Execute Visual Hook Split FFmpeg command, for example change -crf 20 for quality or swap libx264 for libx265.
• Add a Google Drive or S3 upload node after the split step to archive every extracted hook.
• Chain a Kling, HeyGen, or other avatar API call directly after Show Hook Download Link to automatically produce the avatar reaction variant.
• Increase the wait durations if you process longer videos that take more time on RenderIO.