Browse categories
Explore
Fiverr Pro
English
$
USD
I will build custom high quality video datasets for ai training
AI Dataset Expert LoRA Data Engineer AI Image Specialist Dataset Creator
Hi! I specialize in building high-quality, custom image datasets for AI model training (Stable Diffusion, Flux, Wan Video LoRA). >
I know that data collection, cleaning, and captioning are the most ti...
About this Gig
Tired of manually searching for clean, non-watermarked video clips?
Building massive datasets for AI training or finding specific B-roll takes forever. I use a custom-built AI pipeline to source, filter, and process hundreds of high-quality clips tailored to your exact prompt.
What My AI Pipeline Does:
- Deep Sourcing: Extracts raw videos based on your specific text prompts.
- Zero-Garbage (OCR & CLIP): Scans every frame to reject clips with stock watermarks (Shutterstock, etc.), text, or irrelevant subjects.
- GPU Scene Detection: Cuts long videos perfectly at scene changes (delivering ideal 2-12s single-shot clips). No awkward mid-action cuts!
- YOLO Cropping: Uses AI to track the main subject and auto-crop horizontal videos into 9:16 vertical Shorts/Reels.
- AI-Ready Structuring: Generates .txt captions and metadata.csv for Kohya, Finetrainers, or OneTrainer.
Perfect For:
- AI Developers (Text-to-Video models).
- YouTube Creators (Niche background B-roll).
You Get: Perfectly cut .mp4 files in a ZIP archive.
️ IMPORTANT: Please message me BEFORE ordering!
FAQ
Do you provide metadata for AI training?
Yes! In the Premium package, I deliver a detailed metadata.csv file and individual .txt captions for each video.
Can you crop horizontal videos to vertical 9:16?
Absolutely. I use YOLO AI object detection to track the main subject and automatically crop horizontal footage into perfect 9:16 vertical videos.
Do you offer background removal?
Yes, it's available as a Gig Extra. I use Rembg AI to process the video frame-by-frame, delivering transparent footage.
