Free local Ideogram 4 pipeline: ComfyUI + LM Studio + Qwen 3.6 on RTX 3090

Melvin Vivas · X post · 2026-06-20 · Open on X

Topics: AI Dev Tools & Productivity, Prompt & Context Engineering · Level: intermediate

Summary

The creator runs Ideogram 4 locally on an RTX 3090 for free image generation. Inside a ComfyUI workflow, custom LM Studio nodes call the qwen3.6-35b-a3b-mtp model to turn a simple prompt into a JSON prompt, which then goes to Ideogram 4.

Key points

• Hardware: Ideogram 4 runs well on a single RTX 3090 GPU (24GB VRAM).
• Pipeline: simple prompt -> LM Studio -> qwen3.6-35b-a3b-mtp -> Ideogram 4, all inside ComfyUI.
• Custom LM Studio nodes were added to the ComfyUI workflow template to create JSON prompts.
• Using an LLM to expand a short prompt into a structured JSON prompt improves the image model's input.
• The whole setup is local and free.

Resources mentioned

• ComfyUI · tool · x.com · free
  The official X account for ComfyUI, the node-based tool for generative AI workflows, which announces things like ComfyUI Agent.
  Also in: ComfyUI Agent: early access announcement (Melvin Vivas on X · notes), Comfy Router: One API for Image, Video, 3D and Audio Model Providers (Melvin Vivas on X · notes), Generating AI Video Locally with MiniMax H3 in ComfyUI on an RTX 3090 (Melvin Vivas on X · notes), Running MiniMax M3 Locally in ComfyUI on an RTX 3090 (Melvin Vivas on X · notes) and 23 more
• LM Studio · tool · x.com · free
  Desktop app for downloading and running LLMs locally, with a developer mode that serves models through an API.
  Also in: Running LLMs Locally Without an Expensive Rig (Melvin Vivas on X · notes), Adding Vercel AI Gateway as a provider in AIBackends with Devin (Melvin Vivas on X · notes), LoRA Fine-Tune Qwen3.5-2B on Your Tweets with Unsloth Studio (Melvin Vivas on X · notes), AIBackends: An API Layer Between Your App and AI Models (Now with Jev) (Melvin Vivas on X · notes) and 31 more
• Qwen 3.6 35B (MTP) · tool · huggingface.co · free
  Qwen 3.6 mixture-of-experts model (35B total, about 3B active parameters) with multi-token prediction for local inference.
  Also in: Use a Local Model for Confidential Data with Your Agent (Melvin Vivas on X · notes), Running Qwen 3.6 35B locally on an RTX 3090 for agent tool calling (Melvin Vivas on X · notes), Models That Work With the Hermes Agent for Personal Productivity (Melvin Vivas on X · notes), Run Hermes Agent locally with Qwen 3.6 35B MTP in LM Studio (Melvin Vivas on X · notes) and 4 more
• Ideogram 4 · tool · ideogram.ai · free
  Ideogram's image generation model, which works well with structured JSON prompts.
  Also in: Use local Qwen 3.6 to write JSON prompts for Ideogram 4 (Melvin Vivas on X · notes)

Try this

• Add LM Studio custom nodes to the ComfyUI Ideogram 4 workflow template so a local Qwen 3.6 model generates JSON prompts.
• A fully local text-to-image pipeline where a local LLM expands prompts into JSON before Ideogram 4 renders them.