PRAVEEN®

Notes · 003 published · 2 in the works

Field notes.

What I learn, written down properly — step-by-step tutorials with commands you can copy and run, and honest numbers from my own hardware.

  1. LatestVideo models

    Run MiniMax H3 locally on an RTX 5090 or RTX PRO 6000

    Pick your GPU, download the right files, run it in ComfyUI — plus a ready workflow that takes reference images, a video and your own LoRAs.

    № 00324 Sept 20264 min read

  2. Image models

    Run Qwen-Image 2.1 locally on an RTX 5090 or RTX PRO 6000

    The new 7B Qwen-Image: native 2K, transparent PNGs and reference editing in one checkpoint. Pick your GPU, download, run it in ComfyUI.

    № 00223 Sept 20263 min read

  3. Image models

    Run FLUX.2 [dev] locally on an RTX 5090 or RTX PRO 6000

    Pick your GPU, grab the right weights and generate in ComfyUI — multi-reference editing included, plus an 8-step turbo switch.

    № 00122 Sept 20263 min read

In the works

  • DraftingAI Infra

    Serve an open LLM with vLLM on an RTX PRO 6000

    One 96 GB card as a fast, OpenAI-compatible endpoint — install, quantise, benchmark.

  • DraftingTraining

    Train a FLUX.2 [klein] LoRA on one RTX 5090

    Dataset, captions, settings and the time it takes — start to finish on a single card.