After weeks of iteration, I built a system that generates UGC-style product videos of ANY duration with ZERO hallucinations. Here's the magic: 1. Upload your product photo 2. Pick a persona (from database, upload custom, or let AI create one) 3. Set your duration — 5 seconds or 5 minutes, doesn't matter 4. Hit go The system handles EVERYTHING: ✅ Consistent persona across the entire video (no face morphing!) ✅ Product stays intact (no logo changes, no weird mutations) ✅ Seamless transitions between segments (hard cuts that feel like one continuous shot) ✅ Auto-generated captions with karaoke-style highlighting ✅ Smart music selection based on product type ✅ Full control over every parameter The catch? It takes time. A 40-second video = ~15 minutes of processing. But it's fully hands-off — you set it and forget it. Slowly but surely, frame by frame, segment by segment... it just works. Happy to share more details if anyone's interested !

Hey everyone, I’ve been working on an ambitious automation system that generates AI-driven UGC product videos with custom durations. For now, I’m focusing on the proof of concept (POC) using two 10-second clips (total ~20s). Here’s how the pipeline works 👇 1️⃣ Prepare Inputs A submission form captures the product image. The image is analyzed by GPT-4o. Results are stored in Google Sheets with three main fields: ICP, Product Features, Video Settings (can also be manually filled) 2️⃣ Generate First Video (Video 1) An AI agent creates the video prompt. The prompt is sent to the Kie.ai API (Sora 2 model) for generation (following Nate’s tutorial). Output → Video1. 3️⃣ Transcription Used Whisper API to extract the transcription from Video1. The text is then added to Google Sheets for further processing. 4️⃣ Quality Check (Optional) A Gemini-powered row checks for abnormalities or hallucinations in the generated content. 5️⃣ Extract Persona & Last Frame Used Shotstack API to capture 10 frames from Video1. Each frame is analyzed and scored. The system extracts the clearest persona image and last frame automatically. 6️⃣ Prepare for Video 2 Generation The workflow gathers all necessary data (images, prompts, metadata) to build the second video. 7️⃣ Generate Second Video (Video 2) Using this body for the sora-2-image-to-video endpoint: { "model": "sora-2-image-to-video", "input": { "prompt": "{{ $json.output.replace(/\r?\n|\r/g, ' ').replace(/\"/g, '').replace(/[“”]/g, '') }}", "image_urls": [ "{{ $('Limit2').item.json['Last frame'] }}", "{{ $('Limit2').item.json['Persona frame'] }}", "{{ $('Limit2').item.json['Product Photo'] }}" ], "aspect_ratio": "portrait", "n_frames": "10", "remove_watermark": true } } 🎥 Results Attached below: First video (Video 1) Second continuation video (Video 2) Workflow visual 💬 Looking for Feedback I’d love feedback or suggestions on: - Optimizing prompt/body/strategy suggestions. - Improving continuity and persona consistency between videos

Hi everyone, is there anyone working with long format UGC type videos (1-2 minutes) ? What is the best strategy to stitch clips together with smooth video and audio transitions ?

got a 9–5? you don’t need to quit. You need 3–4 focused hours a day. → 1 hour before work → 2 hours after → weekends = 15+ bonus hours That’s 25 hours/week you can build a real business in that. But only if you cut the noise. → no scrolling → no fake “research” → no 6-hour logo design sessions Instead: → pick 1 skill (lead gen, automation, landing pages) → offer it to real businesses → learn + sell every day → get your first client → stack momentum $10K/month isn’t magic. it’s just a few hours a day—used well. 👇 what’s stopping you from starting tonight?

n8n just dropped a new Guardrails node, and it’s a game-changer for anyone building AI automations. In this video, I break down exactly what guardrails are, why they matter, and how you can use them to keep your workflows safe and reliable. Guardrails help you automatically detect and handle things like personal data, secret keys, malicious URLs, or unsafe content before it ever reaches your AI model. I’ll walk you through each guardrail type, show real examples, and explain when and why you’d want to use them. The best part is that you can set all of this up in n8n with no code, so whether you’re protecting your data or building safer AI systems, this video will show you how to do it step by step.