An n8n-powered AI pipeline that turns a plain product photo sent on Telegram into an AI-generated, UGC-style video and uploads it straight to TikTok as a draft.

A plain product screenshot doesn't sell on TikTok. Affiliate videos work best when the product is shown in a real-life scene, but that means filming content, or manually prompting image and video tools, downloading the results, and uploading every video one by one.
This automation reduces the whole process to sending a photo on Telegram. Everything from prompting to uploading runs on its own for my own TikTok affiliate account.
A plain product photo is sent to a private Telegram bot. The bot only accepts my account, uploads valid images to ImgBB, adds them to Google Sheets as queued, and replies with the current queue count.
The bot has several commands that trigger different actions, from generating a single video to processing the whole queue in parallel batches of 5. Telegram sends progress updates along the way.
An OpenAI LLM writes the prompt for a UGC-style scene that showcases the product, and GPT Image generates the first frame from the product photo. The scene style is fully adjustable, so the same workflow can produce any kind of video.
A second LLM prompt directs the motion and camera for the clip, and the first frame is sent to Seedance 2.0 through Kie.ai. The workflow checks the status every 20 seconds until the video is ready and retries failed generations. The prompts are interchangeable, so the scene can be changed per product.
The finished video is sent to Telegram for preview and uploaded through the official TikTok Content Posting API as a draft in the TikTok inbox, ready to publish.
Each product is marked as done or failed in Google Sheets. Errors are handled globally and reported straight to Telegram, so nothing fails silently.
A hands-off pipeline that turns a single product photo into a ready-to-publish, UGC-style affiliate video. Posting 2 videos a day costs only around $12 a month, or roughly $0.20 per video.