Step by step notes from building the AI Character Kit. Real settings, tested on an RTX 3090 Ti with 24 GB.
Two honest ways past the 5 second limit in ComfyUI: chaining Wan 2.2 clips from the last frame, and letting the Wan Animate 2 loop follow a long driving video.
Read the guideRent a 24 GB GPU on Vast.ai for ComfyUI without surprise bills: picking a host, the download charge most guides skip, disk sizing, and stop versus destroy.
Read the guideCan you sell what you make with Krea 2, Qwen-Image-Edit 2511, Wan 2.2 and Wan Animate 2? A plain reading of Apache 2.0 and the Krea 2 Community License.
Read the guideRestyle a realistic AI character as anime, a 3D cartoon or cyberpunk with Qwen Image Edit 2511 in ComfyUI, with tested prompts and what keeps her recognizable.
Read the guideWan 2.2 image to video or Wan Animate 2? What each one needs, what it controls, how long it takes on 24 GB, and a simple rule for picking the right one.
Read the guideA 15 minute recipe for a TikTok-style product ad built only from media you own: a fictional character, a driving clip you filmed or generated, captions and an AI label.
Read the guideWhy Wan 2.2 image to video clips warp, melt faces, freeze or come out blurry in ComfyUI, and the fixes that work, in the order worth trying them.
Read the guideWhy ComfyUI video wants a 24 GB NVIDIA card, how the RTX 3090, 4090 and 5090 compare, and when renting on Vast.ai beats buying, bandwidth costs included.
Read the guideChange your AI character's outfit with Qwen Image Edit 2511 in ComfyUI and keep her face: the prompt pattern, how to describe clothes, and fixes for drift.
Read the guideHow to copy motion, expressions and lip movement from a driving video onto your own AI character with Wan Animate 2 in ComfyUI. Files, settings, prompts and the product trick.
Read the guideBurn big white captions with a black outline into vertical video using ffmpeg drawtext. Tested commands for single cards, timed cards and two-line text.
Read the guideWrite Wan 2.2 image to video prompts that move naturally: describe the motion, not the person. Tested prompts for smiles, sea breeze, walks and hair flips.
Read the guideTested Wan 2.2 14B image to video settings for 24 GB in ComfyUI: fp8 models, LightX2V 4 step LoRAs, 480 x 832 at 81 frames in about 2 minutes, plus fixes.
Read the guideA slot by slot prompt formula for a fictional AI character: age, hair, unique marks, skin and outfit, so the face holds up in every scene and edit later.
Read the guidePlan 30 posts for a fictional AI character from one base image: a weekly mix of stills, clips and talking videos, one batch session, and honest AI labels.
Read the guideSet up Qwen-Image-Edit 2511 in ComfyUI with the 4 step Lightning LoRA: which files to download, where they go, the exact node settings, and prompts that keep a face consistent.
Read the guideKrea 2 Turbo settings for realistic AI portraits in ComfyUI: 8 steps, CFG 1, euler/simple, 896 x 1344, the fp8 model and Qwen3-VL 4B text encoder.
Read the guideA practical ComfyUI method for a consistent AI character: make one strong base image, then edit it into new scenes with Qwen Image Edit 2511. No dataset, no training run.
Read the guide