You'll Never Write an AI Video NSFW Prompt Again — LTX 2.3 + Gemma 4 Auto-Prompt (ComfyUI)

You'll never write an AI video prompt again. In this tutorial I set up PromptNodethingLD —
a free custom ComfyUI node that uses Gemma 4 to AUTO-WRITE camera-locked cinematic prompts for
LTX 2.3 — plus the full "flicker fix" workflow that cleans the output with RIFE interpolation
and RTX Video Super Resolution. Zero prompt skills needed.

Instead of writing a prompt once and hoping, the node auto-prompts in three passes:
• Decide — a fast JSON-forced pre-pass locks a per-beat camera plan (no drift)
• Write — Gemma 4 writes against the locked plan, word budget scales with beat count
• Check + Fix — regex lint catches teleport/hype words, camera restatements and style
violations, then auto-repairs in one surgical pass

Download workflow (Paid) : https://www.patreon.com/aitotts/posts/ltx-2-3-now-its-161340431?utm_medium=clipboard_copy&utm_source=copyLink&utm_campaign=postshare_creator&utm_content=join_link

🧠 GEMMA 4 AUTO-PROMPT
• Why Gemma 4 26B is the recommended model (best quality at 24 GB VRAM)
• How the node feeds Gemma 4 a locked camera plan instead of a raw prompt
• JSON-forced calls keep Gemma's output structured and on-style
• Per-style system rules + negative prompts so Gemma stays inside each look
• Qwen also supported for lower VRAM

🎬 LTX 2.3 + FLICKER FIX
• LTX-2.3 guide rules baked in (main action first, one action per beat, explicit lighting)
• Per-style camera palettes + JUMPCUT mode
• Camera LoRA hints
• Flicker-fix chain: RIFE frame interpolation + RTX Video Super Resolution + upscaling

⬇️ WORKFLOW + NODE (free)
Node + repo: https://github.com/verqqa/PromptNodethingLD
Workflow file (flicker fix): linked in the repo

🔧 REQUIREMENTS
ComfyUI: https://github.com/comfyanonymous/ComfyUI
ComfyUI Manager: https://github.com/ltdrdata/ComfyUI-Manager
llama.cpp: https://github.com/ggerganov/llama.cpp
Model (Gemma 4 26B GGUF): https://huggingface.co/nohurry/gemma-4-26B-A4B-it-heretic-GUFF

💾 VRAM GUIDE
16 GB → q4_k_s (~13 GB)
24 GB → q4_k_m (~15 GB, recommended)

If this helped, drop a like and subscribe — more ComfyUI + AI video workflows coming.

#ltx2 #LTXVideo #comfyui #gemma4 #aivideo #stablediffusion #aitools #aiart

Видео You'll Never Write an AI Video NSFW Prompt Again — LTX 2.3 + Gemma 4 Auto-Prompt (ComfyUI) канала Aitotts