# AI content workflow prompt pack Prompts from my webinar "AI for Content Workflows and Video Editing" (July 9th). Paste them into Claude Code or Codex and change the topics, budgets and file names to match your own work. --- ## 1. Find trending ideas (dynamic workflow + Apify) Needs: Claude Code with Apify connected. This can run for 15 to 60 minutes and uses a lot of tokens (I recommend the $100 to $200 plans). ``` Can you please use a dynamic workflow to check the trending topics on AI on different subjects? I want edge AI, and I also want business cases with AI and any other thing that you can think of. I want you to use four different subagents, and each subagent should run an Apify scraper. You get a budget of fifty cents per agent, and then I want you to report back to me with content ideas for the day based on the trending topics per category. ``` Swap "AI", "edge AI" and "business cases with AI" for your own niche (HVAC, recruitment, marketing agencies...). You can also give each sub-agent one platform: X, YouTube, TikTok, Facebook. --- ## 2. Build your own content atomizer I do not share my skill, but here is how I built it: give Claude Code a content framework from a book, then feed it content. ``` Take this framework. Now I'm gonna give you pieces of content, and I want you to atomize them according to this framework, and then just push them to Notion. ``` What my atomizer does: reads the content, extracts only useful reusable items, classifies them, deduplicates them, previews each item and saves them to Notion. Each atom is tagged: - **Hook**: goes at the top of a post - **Retain**: the middle of a post - **Reward**: the end of a post Framework I used: *100M Leads* by Alex Hormozi. Also recommended: *Influence* by Robert Cialdini. Good inputs: your sales call transcripts, meetings, podcasts, other people's videos. ### Pull a YouTube video to atomize (yt-dlp) Tell Claude Code you want to use the yt-dlp Python library, then: ``` Here is the video by [creator]. Please get the transcript from YouTube. Please also get the comments. ``` The comments are where people ask for new things, so they give you even more ideas. Before reusing an atom, add your own story or take. --- ## 3. Understand a video: transcript vs watching **Transcript route (precise timestamps, for editing):** ``` Please use ElevenLabs to transcribe the first video that's a bit longer, and then tell me what that video is about. ``` **Watching route (Gemini sees picture and sound):** ``` Please use Gemini 3.1 Flash Lite to check the shortest video in there, and tell me what it's about. ``` **Pre-publish safety check (my use case):** have Gemini watch the video before you publish so it flags any API key, ID or email shown on screen and tells you which time range to blur. --- ## 4. Code-based editing with HyperFrames (about one cent per short) Needs: an ElevenLabs transcript and HyperFrames available to Claude Code. Best for talking-head videos. ``` Can you now use Hyperframes to edit this video? My objective is to add science icons and things whenever I talk about microscopes, atoms, quantum physics. You can also start to add the captions at the bottom of the screen. Highlight the word that's being said. And if you can, also add a quick two-second introduction explaining what the video's gonna be about before the actual main video starts running. ``` Replace the topic and icons with your own. Follow-ups you can ask for: put things on the left or right, zoom your face in or out, put your face in a corner with another image behind. --- ## 5. Generative editing with Gemini Omni (about 7 cents per second) Needs: a Gemini API key. Gemini Omni caps at about ten seconds per clip, is not available in Europe or the UK without a VPN, and may refuse to turn a real person into someone else. Changing your T-shirt or background is the fallback I suggested. ``` Can you please use Gemini Omni Flash with my Gemini API key to take that video, and whenever that guy is snapping his fingers, you switch him into a different person. Check out what he says and find the most appropriate persona to change it. I want you to generate four different options for this video. ``` Budget for three or four runs per clip to get a result you like. --- ## Which editing route to use | | Code-based (HyperFrames) | Generative (Gemini Omni) | |---|---|---| | Input | Timestamped transcript (ElevenLabs) | The video itself (Gemini) | | What it does | Adds cards, icons and captions on top of your footage | Regenerates the video | | Cost | You pay for the transcript, about a cent | About 7 cents per second, several runs | | Best for | Talking-head videos | Creative effects: new background, new look |