Meesa Multis Maker

The ultimate tool for creating Ultima Online multi structures and GUMP graphics with AI-powered texture regeneration

Features � Installation � Getting Started � Editors � AI Integration � Shortcuts � License

---

Overview

Meesa Multis Maker is a Windows desktop application for composing, editing, exporting, and AI-regenerating isometric game assets used in Ultima Online-style multis and GUMPs. It provides a powerful canvas for placing item art on an isometric diamond grid with support for complex transforms and integrates AI pipelines for texture and static regeneration.

System Requirements

• OS: Windows 7 or later
• .NET Framework: 4.8.1 or later
• Ultima Online Client: Required for MUL/UOP asset files
• GPU (Optional): NVIDIA GPU recommended for local model inference (LLM, vision, audio gates) and local ComfyUI
• ComfyUI Server (Optional): For server-based AI image generation
• Python 3.12 + ffmpeg (Optional): Audio quality gates (vocal bouncer, CLAP match, captions) via pip install faster-whisper laion-clap transformers librosa torch torchvision
• llama.cpp server binary (Optional): llama-server.exe beside its ggml-*.dll files in MeesaMultisMaker/LLM/bin/ for local LLM/vision features (weights auto-download)

---

Features

Isometric Canvas Editor (Main Form)

• Diamond grid system - Standard 44x44 pixel Ultima Online tile format
• Multi-object selection - Ctrl+Click for multi-select, marquee selection with Ctrl+Drag
• Z-ordering and layering - Full control over object depth and draw order
• Pan & Zoom - Middle-click pan, Ctrl+Scroll to zoom (0.2x to 4x)
• Snap-to-grid - Objects automatically align to isometric grid positions
• Z-level filtering - Filter visible objects by Z height range

Transform Tools

• Scale - Resize objects (0.1x to 10x) with +/- keys
• Rotate - Free rotation with R key, 15 degree or 90 degree snap with Shift/Ctrl
• Flip - Horizontal (F) and Vertical (Shift+F) flipping
• Skew/Distort - Corner-based perspective transform (K key)
• Pixel Offset - Fine-tune positioning with Alt+Arrow keys

Slice Tool

Create vertical slices of placed images based on selected diamond tiles - perfect for creating building pieces that align perfectly with the isometric grid.

Lock System

• Lock objects to prevent accidental editing
• Separate locked/unlocked object lists
• Multi-select lock/unlock operations

AI Integration (ComfyUI)

All image generation runs through a ComfyUI server (local or remote):

• More model options - Use any Stable Diffusion checkpoint, including Flux.2 Klein (flux-2-klein-9b.safetensors + flux2 CLIP + flux2-vae), which gets dedicated inpaint workflows when selected
• Advanced workflows - Inpainting, ControlNet, and more
• Remote processing - Can run on a separate machine or cloud

AI Features:

• Image-to-Image - Transform existing sprites with AI
• Inpainting - Selective region regeneration with masks
• Text-to-Image - Generate new assets from text prompts
• Batch Processing - Regenerate multiple selected objects at once
• Old/New Toggle - Compare AI-generated results with originals
• Drop Black Pixels - Automatic background removal (adjustable post-generation)

Import/Export

• Canvas Export - Save entire canvas as PNG
• Multi Text Import/Export - Load/save Ultima Online multi format files
• Parts Export - Export individual components for AI processing
• MUL File Export - Save directly to Ultima Online asset files

Settings

---

Main Editors

Painter Form

A full-featured layer-based pixel art editor for creating custom sprites.

Features:

• Multiple canvas sizes - 512x512, 768x768, 1024x1024, 1536x1536
• Layer system - Add, delete, reorder, rename, lock, and adjust opacity
• Drawing tools:
  • Brush (soft round with anti-aliasing)
  • Pencil (hard pixel-perfect)
  • Eraser
  • Flood fill
  • Line tool
  • Rectangle tool
  • Circle/ellipse tool
• Fill shapes - Toggle between outline and filled shapes
• Mask mode - Paint masks for AI inpainting
• Zoom - Ctrl+Scroll or Ctrl+/Ctrl- to zoom
• Undo - Per-layer undo history (Ctrl+Z)
• Clipboard - Paste images directly (Ctrl+V)
• Image effects panel - Brightness, contrast, hue, saturation, pixelize, noise, dither
• Send to Canvas - Export merged image to main form
• Send to GUMP Editor - Export to GUMP Editor

GUMP Editor

Specialized editor for Ultima Online UI graphics (GUMPs).

Features:

• GUMP palette browser - Search and filter available GUMPs
• Drag-and-drop placement - Add GUMPs to canvas
• Multi-selection - Select and transform multiple GUMPs
• Transform tools:
  • Scale (resize with handles or S+/S- buttons)
  • Rotate (15� increments or free rotation)
  • Flip Horizontal / Flip Vertical
  • Skew (adjustable X/Y skew with live preview)
  • Slice (divide gumps into grid pieces)
• Layer management - Organize GUMPs in layers
• AI regeneration (ComfyUI):
  • Single GUMP regeneration
  • Multi-GUMP combined regeneration (processes as one image, splits result)
  • Alpha mask preservation
  • Post-generation black pixel adjustment - Toggle and adjust threshold after AI generation
  • Old/New comparison toggle
  • Revert to original
• Custom image import - Add PNG images as GUMPs
• Canvas export - Save compositions as PNG

MUL Viewer

Browser for viewing and generating multi structures from MUL files.

Features:

• Multi entry browser - List all multis with dimensions
• Parts view - See individual components with TileID, X, Y, Z, Flags
• Preview rendering - Isometric preview with pan and zoom
• Hide empty entries - Filter out empty multi slots
• Normalize coordinates - Center multi at origin
• Export text format - Copy multi definition to clipboard
• Send to Canvas - Import selected multi to main editor
• Add to Canvas - Append multi to existing objects

Structure Generation (AI-like):

• Select one or more multis as training data
• Configure output dimensions (width x height)
• Adjust generation parameters:
  • Kernel size
  • Epochs
  • Learning rate
  • Similarity factor
• Choose algorithm:
  • Neural network (SimpleCnn)
  • Wave Function Collapse (WFC)
• Generate new structures based on learned patterns

3D Editor

Text-to-3D sprite pipeline ("MultisMaker1"): prompt ComfyUI for a model, orbit it through the 8 UO facings, slice it into floor-aligned UO tiles, and send the tiles to the canvas.

• Text to PLY - prompt, seed, sampler steps via ComfyUI; optional init image (including straight from the Painter); Use view as input for iterative refine; timestamped .ply outputs
• Viewport - orbit/zoom, 8 UO facing snap + Set S, UO-mode pitch lock (~26.57), splat-density cap for previews, middle-drag pan on ground plane, Quad Top/Front/Side/Persp views, direction buttons
• Tile slicing - floor-space hard-cut by splat center (no blend); TILE 44x44, Z_PIXEL 4, FloorGridCells 10; floor grid drives tiles (no manual X/Z counts); origin-relative send, canvas centers batch; calibrated 1 cell = 44px export + zoom; centered crop; dials-baked + slice-time yaw; orbit re-cuts, dials re-cut debounced, EnsureFreshSlice on send/export
• Tile dock - slice/clear, isolate render path, send-all (TilesReady), export-all, 9-slice expand + hollow, auto-refresh
• Edit survival - tile list survives delete/clone/undo via Reclassify / RefreshSlicesKeepActive; undo split IsDelete vs mutate
• Dials + offsets - rotation/scale bake into positions; Scale X/Y/Z lock (default on); OffX/OffY numerics + O0 mirror PixelOffsetX/Y, non-destructive, no bitmap padding
• Loop mode - auto-resubmit forever (fresh image and/or fresh LLM-dreamed prompt each round)
• Spintable export - 24-frame rotation sheets (+ GIF)

Animation Editor

Viewer/player/editor for Ultima Online mobile animations (anim.mul family): scrub frames, edit, and export PNG sequences. See the full guides in docs/: ANIMATION_EDITOR_GUIDE.md, ANIMATION_EDITOR_QUICK_REF.md, ANIMATION_EDITOR_TROUBLESHOOTING.md.

Audio Editor

ComfyUI-powered sound design: text-to-SFX, text-to-music, voice cloning / voice design (Qwen TTS), a webcam-driven ambient Watcher that scores the scene with generative foley + music layers, and UO sound.mul / music-folder integration.

• SFX / Music / Voice tabs - prompt, negative prompt, duration, seed, steps, per-kind workflows (FASTAUDIOTGEN SFX, WORKINGAUDIO Music, Qwen TTS clone/design); category forced SFX/Music; reprompt off by default; saver filename_prefix pinned to short IDs (no Errno 22); upload/audio then upload/image fallback; queue/history polling with fail-fast server errors
• Filenames + ledgers - short IDs (yyMMdd_HHmmss_kind), never prompt text; master pipe-delimited AudioPrompts.csv (id|created|kind|seed|duration_s|audio_file|prompt) + library.jsonl with prompt/seed/workflow/file/scene/caption
• Waveform + edits - transport, trim/fade/normalize/gain/resample/mono/reverse/silence-crop/loop-crossfade; edge-silence trim (-45dB + pad + micro-fades); ffmpeg decode to editable WAV + mm_savewav sibling
• Radio - AutoQueue keeps 1 ahead, tight 750ms end-poll swap; ON AIR mini-views with playheads
• Watcher sensing + scheduling - webcam via ffmpeg dshow, VL Qwen2-VL-2B via local server; two editable persisted questions (foley/music); sense every N sec (never blocks), 1s scheduler prefetches before clip end reusing freshest prompt; foley length = period, music overlap + 15s minimum hold; adaptive render estimates with EWMA; queue-to-ready timing; 15s failure backoff; layer-disable mid-render keeps file without airing; snapshot force-queues
• Change-gating - 16x16 gray-hash frame gate (adjustable Δ%, default 4%) + word-Jaccard prompt gate (0.85); static scenes loop on-air file with zero GPU (loop-prepped CrossfadeLoop), dynamic scenes prefetch with lead time
• Crossfaded swaps - dual-alias MCI flip-flop, 1s foley / 2s music over 12 steps (clamped to 60% body), late renders degrade to fade-in, cancel finishes instantly
• DJ timeline - vertical cards newest-first (orange foley / cyan music), time/layer/dur/seed header, clip + prompt, live progress, ON AIR pulse, played Xs stamp; click for full PROMPT/SCENE/HEARD, hover tooltip, double-click loads to Clips, Clear, 200 cap
• Quality gates (optional Python sidecar AudioGateServer.py on 127.0.0.1:18171, killed on close) - Whisper vocal bouncer (≥4 words on instrumental intent → one hardened retry, air best, loser kept); CLAP match floor (reject <0.10 → one retry, air best); LUFS match (-19 music / -16 foley, TP -2, LRA 11, two-pass loudnorm with rate pin); scores in log + timeline/library tags; Bounce vocals / CLAP gate checkboxes
• Captions (laion/sound-effect-captioning-whisper, lazy load, first 30s) - auto-caption aired + manual/radio takes into library Caption (searchable, backfilled by Describe); Describe buttons on timeline cards + clip panel
• Manual/radio behavior - same trim + loudnorm; transcribe/CLAP measured + logged (no auto-retry); radio buffers, library appends
• UO integration - legacy pair + soundLegacyMUL.uop (1655 entries, in-place fixed-size replace + .bak, FitToFrames); Music/Digital/*.mp3 + Config.txt browser; slot play, clip-to-slot, save, fit; 22050 mono conversion for legacy slots
• Library tab - prompt/kind/scene/caption search, missing-file flags, prune, reuse prompt, play

Image Editor

Quick image effects editor for adjusting brightness, contrast, and applying filters to selected objects.

Map Viewer

World map browser with tile/static editing and AI replacement.

Features:

• Facet selection - Browse Felucca, Trammel, Ilshenar, Malas, Tokuno, Ter Mur
• Pan and zoom - Navigate large maps efficiently
• Minimap generation - Overview of entire map
• Coordinate navigation - Jump to specific X,Y location
• Selection modes:
  • Land tiles
  • Statics
  • Lasso selection for irregular areas
• Z-level filtering - Show statics within Z range
• Static manipulation:
  • Move with numpad keys (X, Y, Z)
  • PageUp/PageDown for Z adjustment
• AI tile replacement (ComfyUI):
  • Replace mode - Simple regeneration
  • Context mode - Context-aware inpainting
  • Batch processing of selected tiles
• Show statics toggle - View map with or without statics
• TileData display - View item properties

Export Map

Export the currently loaded map as a high-resolution isometric image or a set of tiled JPEG files with an interactive HTML viewer.

Export Options:

• Zoom (px/tile) - Output resolution from 1 (tiny overview) to 44 (full detail per tile)
• Include Static Objects - Render statics on top of land tiles with hue, translucency, and partial-hue support
• Export HTML & Minimap Only - Regenerate the viewer, minimap, and globe texture without re-rendering tiles (useful for quick iteration)
• Generate Mipmap Only - Rebuild the mipmap pyramid from existing tiles without re-rendering
• Region selection - Export the entire map or a custom rectangular region (Start X/Y, End X/Y)

Output Modes:

Tiled Export Contents:

• tile_R_C.jpg - Full-resolution rendered tiles (row R, column C)
• viewer.html - Interactive pan/zoom map viewer (open in any browser)
• minimap.jpg - Isometric radar-color overview for quick navigation
• globe_texture.png - Composited image from the coarsest mipmap level, suitable for projecting onto a 3D globe
• L1/, L2/, � LN/ - Mipmap tile pyramids for efficient zoomed-out viewing

HTML Viewer Features:

• Drag to pan, mouse-wheel to zoom
• Clickable minimap for quick navigation
• Automatic mipmap level switching when zoomed out
• Only visible tiles are loaded into the DOM for memory efficiency
• Zoom buttons: +, ?, 1:1, Fit

Rendering Details:

• Tiles are rendered using textured isometric projection with deformed terrain (Z-height per corner)
• NoDraw land tiles and statics are automatically filtered out
• Statics are drawn with correct depth sorting (painter's algorithm), hue coloring, translucency, and foliage layering
• Tiles are rendered in parallel across all CPU cores for maximum speed
• Art assets are pre-loaded and serialized to byte arrays so parallel threads never contend on shared GDI+ objects

---

Screenshots Gallery

---

Installation

Download

1. Download the latest release from the Releases page
2. Extract the ZIP file to a folder of your choice
3. Run MeesaMultisMaker.exe

First Launch

1. Accept the license agreement
2. Browse to your Ultima Online installation folder when prompted
   • The application will search common paths automatically:
     • C:\Program Files (x86)\Electronic Arts\Ultima Online Classic
     • C:\Program Files (x86)\UOForever\UO
     • And other common locations

Supported Asset Formats

• MUL Files - Standard OSI format (art.mul + artidx.mul)
• UOP Files - Enhanced format (artLegacyMUL.uop)
• PNG Art Folders - UOFItems-style PNG directories

---

Getting Started

Basic Workflow

1. Load Assets - Click "MUL Folder..." to select your Ultima Online installation
2. Browse Palette - Use the left panel to search and browse available art
3. Place Objects - Drag items from the palette onto the canvas
4. Transform - Select objects and use transform tools to position them
5. Export - Save your creation using the Export button

Creating a Multi Structure

1. Set grid size using Width/Height controls (e.g., 10x10)
2. Click "Create Grid" to initialize the canvas
3. Drag foundation pieces from the palette
4. Add walls, set appropriate Z levels with Z+/Z- buttons
5. Layer decorations using L+/L- for draw order
6. Export as Multi Text for use in-game

---

AI Integration

All image generation runs through a ComfyUI server (local or remote). Selecting a flux-2-klein checkpoint switches inpaint workflows to the dedicated Flux.2 Klein graph (Flux UNET + flux2 CLIP + Flux2 VAE).

ComfyUI (Server-based)

ComfyUI provides more flexibility with model selection and advanced workflows.

Advantages:

• ? Wide variety of checkpoint models
• ? Advanced features (ControlNet, inpainting, etc.)
• ? Can run on a remote server
• ? More control over generation parameters

Requirements:

• Running ComfyUI server (local or remote)
• Stable Diffusion checkpoint models

Setting Up ComfyUI

Option 1: Google Colab (Free, No GPU Required!)

ComfyUI can run in Google Colab notebooks for free using their T4 free GPUs.

Step 1. Go to https://colab.research.google.com/

Step 2. Click "New notebook" button.

Step 3. Paste this script into the notebook cell:

!git clone https://github.com/comfyanonymous/ComfyUI.git
!pip install -r /content/ComfyUI/requirements.txt
!pip install pinggy
!mkdir -p models/checkpoints
!wget -O models/checkpoints/sd_xl_base_1.0.safetensors \
https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/resolve/main/sd_xl_base_1.0.safetensors?download=true

!wget -O /content/ComfyUI/models/checkpoints/dreamshaper_8.safetensors \
https://huggingface.co/jzli/DreamShaper-8/resolve/main/dreamshaper_8.safetensors

import pinggy

tunnel = pinggy.start_tunnel(forwardto="localhost:8188")

print("Public URL(s):")
for url in tunnel.urls:
    print(url)

!cd /content/ComfyUI && python main.py --listen 0.0.0.0

Step 4. Click the Play button to run the script.

Step 5. Copy the pinggy URL and paste it into the ComfyUI URL field. Add :80 to the end.

Example: http://eyzzg-33-125-104-65.a.free.pinggy.link:80

Option 2: Run Your Own ComfyUI Server (Local)

1. Install ComfyUI

   git clone https://github.com/comfyanonymous/ComfyUI
   cd ComfyUI
   pip install -r requirements.txt
   

2. Download Required Models Place these in your ComfyUI models/checkpoints folder:

   • DreamShaper 8 - dreamshaper_8.safetensors
   • SD 1.5 - v1-5-pruned-emaonly.ckpt

3. Start ComfyUI

   python main.py --listen 0.0.0.0 --port 8080
   

4. Configure in Meesa Multis Maker

   • Select "ComfyUI (Server)" from AI Backend dropdown
   • Set ComfyUI URL to http://localhost:8080

AI Settings Panel Options

Recommended Generation Settings

Prompting Tips

Good prompts for UO-style assets:

• "isometric game tile, pixel art style, [object description], clean background"
• "3d model isometric view, [object], black background, studio lighting"
• "ultima online style, medieval fantasy, [object]"

Negative prompt suggestions:

• "blurry, low quality, distorted, text, watermark, signature"
• "modern, sci-fi, photo-realistic" (if you want stylized results)

---

Keyboard Shortcuts

Canvas Navigation

Selection

Object Manipulation

Transforms

Editing

Map Viewer (Statics Selected)

Painter

GUMP Editor

---

Configuration

Settings are stored in:

%APPDATA%\MeesaMultisMaker\config.xml

Audio generations live in Documents\MeesaMultisMaker\Audio\ (library.jsonl prompt ledger, AudioPrompts.csv pipe-delimited prompt log, rendered clips).

Configurable Options

• MUL folder path
• PNG art folder path
• ComfyUI server URL
• Default AI generation parameters (prompt, steps, CFG, denoise, sampler, scheduler, resolution)
• Audio watcher cadence, scene-change threshold, vocal bouncer, CLAP gate
• Update check preferences
• Window position preferences

---

Troubleshooting

Common Issues

"Parameter is not valid" error on images

• This can occur if image data becomes corrupted. Try reloading the application.
• Ensure your MUL files are not corrupted.

ComfyUI connection fails

• Verify ComfyUI is running and accessible
• Check firewall settings
• Ensure the URL includes the port (e.g., http://localhost:8080)
• For Google Colab, make sure to add :80 to the pinggy URL
• Test with "Test Connection" button

Missing art/textures

• Verify the MUL folder path in settings
• Check that art.mul and artidx.mul exist (or artLegacyMUL.uop for UOP format)

AI generation produces black images

• Enable "Drop Black Pixels" option
• Adjust the "Black Threshold" slider after generation
• For ComfyUI: Ensure checkpoint models are installed

AI output is wrong size (too large/small)

• The AI automatically resizes output to match the original image dimensions
• If using scaled gumps, the output will match the source image size, not the displayed size

Multi generation produces empty results

• Select at least one multi entry as training data
• Ensure selected multis have components
• Try adjusting kernel size and epochs

Google Colab session disconnects

• Free Colab sessions have time limits and will disconnect after inactivity
• Simply re-run the script to start a new session
• You'll get a new pinggy URL each time

Getting Help

• GitHub Issues: Report bugs or request features

---

File Formats

Multi Text Format

Tab-separated values with columns:

TileID  X   Y   Z   Flags
0x0001  0   0   0   1
0x0002  1   0   0   1

Supported Image Formats

• PNG (recommended for transparency)
• BMP
• JPG/JPEG
• GIF

MUL Files Used

• art.mul / artidx.mul - Item and land art
• artLegacyMUL.uop - UOP format art
• gumpart.mul / gumpidx.mul - UI graphics
• gumpartLegacyMUL.uop - UOP format gumps
• map*.mul - World map data
• statics*.mul / staidx*.mul - Static objects
• multi.mul / multi.idx - Multi structures
• tiledata.mul - Item properties

---

Credits

• Developer: MeesaJarJar
• ComfyUI: comfyanonymous
• Stable Diffusion: Stability AI
• DreamShaper: Lykon

---

Building from Source

Requirements: Windows, .NET SDK (builds the .NET Framework 4.8.1 target), and NuGet restore (automatic).

git clone https://github.com/MeesaJarJar/MeesaMultisMaker.git
cd MeesaMultisMaker
dotnet build MeesaMultisMaker/MeesaMultisMaker.csproj -c Release

The binary lands in MeesaMultisMaker/bin/Release/MeesaMultisMaker.exe. No game files ship with the repo: point the app at your own Ultima Online installation on first launch.

Feature dependencies (all optional, all downloaded on first use except where noted):

• AI image generation - needs a running ComfyUI server (local or remote); checkpoints (including Flux.2 Klein) live in the server's models/ folders.
• LLM + vision models - GGUF weights auto-download on first use.
• llama.cpp server - llama-server.exe is NOT in the repo: download a Windows release from llama.cpp and place it (with its ggml-*.dll files, already in MeesaMultisMaker/LLM/bin/) next to them.
• Audio quality gates (vocal bouncer, CLAP match, captions) - need Python 3.12, ffmpeg, and pip install faster-whisper laion-clap transformers librosa torch torchvision; without them generation still works, gates report unavailable.
• Push to MeesaJarJar - currently disabled in code (AppConfig.JarJarPushEnabled = false); the buttons render grayed-out. Flip the flag to restore it against your own server.

---

License

This project is licensed under the MIT License. See the LICENSE file for full terms.

---

Contributing

Pull requests are welcome:

• Bug reports via GitHub Issues
• Feature requests and suggestions
• Community tutorials and guides
• Sharing your creations!

---

Made with love for the Ultima Online community

Copyright 2024-2026 MeesaJarJar. All rights reserved.