CivArchive
    ← All articles
    Published June 22, 2025by Rempeyek

    The Holy Grail — Ultimate Modular GIGA Workflow (With or Without Reference)

    1.3K views2 reactions0 comments on CivitAI40 collected
    all-in-onetxt2imgcrimson rift studioworkflowscomfyuifacecrimsonriftstudiogiga workflowface detailerupscalehand detailerimage generateposeworkflowreference

    "The Holy Grail" is a fully modular, cinematic GIGA workflow for ComfyUI, designed to handle every major AI image generation style — with or without face/prompt/picture references.

    This workflow supports:

    ✅ IPAdapter Face Reference (optional)
    ✅ ControlNet Pose Reference (optional)
    ✅ LoRA stacking for character fusion
    ✅ Dynamic prompt structure
    ✅ Multi-stage KSampler rendering
    ✅ High-res detailer and 4x upscaling

    It’s perfect for:

    • Anime-style battle scenes

    • Cinematic poster-like shots

    • Splash-art composition

    • Creative portrait or full-body character renders

    You can create anything — from OC fanart to anime-style cinematic clashes like Skirk vs Primordial Dragon or Aether vs Raiden Shogun.

    ⚙️ 💡 How to Use This Workflow:

    🔀 Step 1: Load the .json file into ComfyUI

    Go to Load > Load (custom nodes) and pick The Holy Grail.json

    🎛️ Step 2: Choose Reference Mode

    ⚠️ This workflow is hybrid. You can choose:

    ModeWhat to Do✅ WITH Face & Pose ReferenceTurn ON all IPAdapter & ControlNet nodes → input image & pose❌ WITHOUT ReferencesBypass those nodes (right click > "Bypass Node") or turn off their switches (e.g., IPAdapter → off)

    You can freely swap:

    • IPAdapter images (for face control)

    • OpenPose (for pose control)

    • LoRA models (for style & outfit)

    ✍️ Step 3: Input Prompt & Negative Prompt

    📥 Prompt Area (in Conditioning node):

    css

    CopyEdit

    masterpiece, anime artstyle, ultra-detailed, cinematic lighting, epic battle, full body

    📥 Negative Prompt:

    scss

    CopyEdit

    (worst quality:2), (low quality:2), blurry, bad anatomy, extra limbs, watermark, ugly

    🛠️ Step 4: Settings & Recommendations

    ParameterValueModel CheckpointmeinamiX_v12, revAnimatedSamplerDPM++ 2M KarrasSteps30–40CFG Scale7.5–8.5Resolution1024x1536 (portrait)Upscale2x–4x UltraSharp built-in

    🧩 Included Modules:

    • ✅ LoRA Stack Loader

    • ✅ Face Restore & Detailer

    • ✅ Dynamic Prompt Conditioning

    • ✅ ControlNet OpenPose input

    • ✅ IPAdapter Advanced

    • ✅ Post-processing upscale / VAE fix

    ✨ Credit:

    👤 Created by Rempeyek (Crimson Rift Studio)
    🔧 Powered by ComfyUI Modular Node System
    📖 Inspired by cinematic anime workflows

    ✅ Notes:

    • To turn OFF references, just bypass the following nodes:

      • IPAdapter (face ref)

      • ControlNet (pose ref)

    • To use your own embedding, LoRA, or VAE, plug into the loader sections.

    • This workflow is beginner friendly but pro-level ready — modify as needed!

    Attachments