Video intake, captioning and auto-posting pipeline

Deliverables

• Drive watch, CloudConvert conversion and Deepgram transcription
• OpenAI content classification with a three-route category router
• Category logging in Google Sheets with duplicate checks
• Caption and hashtag scenario using Apify and OpenAI
• Platform router posting through Metricool
• Crash report sheets and Telegram fail alerts

The challenge

Posting one video to several platforms means watching for new files, converting formats, writing a caption and hashtags for each platform, uploading everywhere and keeping track of what went out. Done by hand, this eats time on every upload, and a failed post on one platform can easily go unnoticed.

Our approach

The intake scenario watches a Google Drive folder, downloads each new video, converts it with CloudConvert and transcribes it with Deepgram. OpenAI analyzes the transcript to classify the content, and a router sends the video down one of three category routes. Each route checks a Google Sheet so a video is not logged twice, adds it to that category's sheet, notifies the team on Telegram, moves the file to the right Drive folder and sends a JSON payload to the next scenario. The caption scenario receives that payload, has GPT draft the angle, runs Apify actors to collect hashtag and trend data, and uses OpenAI to write the final caption and hashtags, saving the result to a sheet before passing it on. The posting scenario parses the payload, downloads the video, runs a CloudConvert job with status checks, uploads the final file to Drive and hands it to a platform router. Its routes schedule posts through Metricool for Facebook, YouTube, TikTok and one more channel, while a separate OpenAI caption step prepares the Threads post. Every route writes a crash report row and sends a Telegram fail signal when something breaks, and error handlers let the scenarios carry on past a failed module.