See llms.txt for all machine-readable content.

Back to Templates

Create Ken Burns zoom video clips from text prompts with Pollinations and FFmpeg

Created by

Created by: Faceless Channel OS || trungcln
Faceless Channel OS

Last update

Last update 2 days ago

Categories

Share

Quick overview

This self-hosted workflow collects a text prompt and clip length, generates a 1080×1920 image via Pollinations, then uses local FFmpeg to create a vertical MP4 with a Ken Burns-style slow zoom and returns the output file paths.

How it works

  1. Receives an image prompt and desired clip length from an n8n form.
  2. Appends cinematic style keywords, URL-encodes the prompt, and builds per-run file paths in a unique /tmp folder.
  3. Creates the temporary output folder on disk.
  4. Downloads the AI image from Pollinations using curl and saves it to the temp folder.
  5. Uses FFmpeg to crop/scale the image to 1080×1920 and render a zoompan animation into an H.264 MP4 of the requested duration.
  6. Checks FFmpeg’s exit code and returns the image path, video path, prompt, and clip length.

Setup

  1. Run this workflow on a self-hosted n8n instance with the Execute Command node enabled (it is not available on n8n Cloud).
  2. Install curl and ffmpeg (with libx264) on the n8n host and ensure both are available on the system PATH.
  3. Ensure n8n is allowed to write to the configured temp location (default /tmp) and adjust the output directory in the code step if your environment restricts filesystem access.

Requirements

  • No API key or account anywhere in this workflow - Pollinations is a free, keyless image API

Customization

  • Change zoom speed or ceiling by editing zoom+0.0012 / 1.15 in the FFmpeg command, or swap in any other keyless image API by editing the URL in the first Code node

Additional info

The zoompan filter's d parameter (frame count) is computed from your requested Clip Length x fps, so the zoom always finishes exactly at the end of the clip regardless of duration - it isn't hardcoded to one fixed length. This template outputs video only, no audio track; pair it with a separate voiceover/TTS template to add sound. Tested end to end: image download and zoom render both verified on a real run before submission.