The Builder's Guide — Workflows for Every Medium

$ ./guide --all-mediums

Every creative medium follows the same fundamental loop: describe → generate → refine → ship. What changes is the toolchain. This guide covers real workflows for each, with emphasis on free and open source tools.

The Universal Prompt Framework

Good prompts share the same anatomy regardless of medium:

[ROLE] You are a [specialist type].
[CONTEXT] I'm building [what] for [purpose].
[CONSTRAINTS] Must be [limitations: size, format, budget, platform].
[OUTPUT] Generate [specific deliverable] in [format].
[ITERATION] I'll ask for revisions — maintain context.

Tips that work everywhere:

• Start broad, then constrain. "Design a lamp" → "Design a minimal desk lamp, 30cm tall, for 3D printing in PLA, with a built-in cable channel"
• Ask for multiple options. "Give me three approaches" reveals possibilities you didn't consider
• Reference existing things. "Similar to X but with Y" is more precise than pure description
• Chain prompts. Break complex projects into stages rather than one massive prompt
• Save your good prompts. Build a personal library — they compound in value

Software & Code

Workflow: Prompt → Working Prototype

1. Architecture first

"I'm building a CLI tool in Python that monitors a directory for new images and auto-optimizes them. Suggest the architecture: what libraries, what file structure, how to handle the file watcher."

2. Generate scaffolding

"Generate the project structure with pyproject.toml, a watcher module using watchdog, and an optimizer module using Pillow. Include type hints and a basic test file."

3. Iterate on components

"The watcher detects files but crashes on permission errors. Add retry logic with exponential backoff and logging."

4. Test and ship

"Write pytest tests for the optimizer module covering JPEG, PNG, and WebP inputs. Include edge cases for corrupted files."

Tools (All Free)

Project Management for Solo Devs

• Git + GitHub — Version everything. Commit often. Use branches for experiments.
• GitHub Issues — Track bugs and features without leaving the terminal (gh issue create)
• Markdown docs — Keep a DECISIONS.md in every project root. Future-you will thank present-you.
• Makefile or Justfile — Codify your build/test/deploy commands. Don't rely on memory.

3D Printing & Physical Prototyping

Workflow: Prompt → Printed Object

1. Describe the object with constraints

"Design a wall-mounted headphone stand for Sony WH-1000XM5. Extends 12cm from wall, supports 400g, includes cable hook underneath. PLA 3D print, minimal style."

2. Generate in OpenSCAD (parametric)

"Write OpenSCAD code for this design. Make all dimensions parametric so I can adjust wall_distance, hook_angle, and mount_hole_spacing."

3. Validate before printing

• Import STL into your slicer — rotate, inspect
• Check wall thickness (minimum 1.2mm for FDM)
• Analyze overhangs (>45° needs supports or redesign)
• Verify dimensions against real measurements

4. Slice and print

• BambuStudio, PrusaSlicer, or Cura — AI-optimized profiles handle most settings
• Print time: 30 min to 24 hours depending on size

5. Iterate with one prompt

"The hook snapped under load. Make it 3mm thicker and add a fillet at the base."

New file. Reprint. Failure-to-fix drops from days to minutes.

Beyond 3D Printing

Generative Art & Design

Workflow: Prompt → Visual Output

1. Concept description

"A cyberpunk cityscape at golden hour, isometric perspective, pixel art style, 1024x1024"

2. Generate with local tools (free)

• ComfyUI — Node-based Stable Diffusion interface. Most powerful free option.
• AUTOMATIC1111 — Web UI for Stable Diffusion. Simpler to start with.
• Invoke AI — Polished open source image generation suite.

3. Refine

• img2img for variations on a good result
• Inpainting to fix specific regions
• ControlNet for structural guidance (pose, depth, edges)
• Upscale with Real-ESRGAN (free, runs locally)

4. Post-process

• GIMP — Free Photoshop alternative, powerful with plugins
• Inkscape — Free vector editor, good for converting raster to SVG
• Krita — Free digital painting app, excellent brush engine

Prompt Engineering for Images

• Subject + Environment + Style + Technical = consistent results
• Negative prompts matter: "no text, no watermarks, no extra fingers"
• Seed locking: find a good composition, lock the seed, vary the prompt
• Batch generate: make 20 variants, pick the best, iterate from there

Organizing Your Visual Assets

• Name files with descriptive slugs: cyberpunk-city-isometric-v3-final.png
• Keep a prompts.md alongside outputs so you can reproduce results
• Use folders by project, not by date
• Back up your ComfyUI workflows as JSON — they're your recipes

Music & Audio

Workflow: Prompt → Track

1. Composition

• Udio — Text-to-music generation, surprisingly good for prototyping ideas (freemium)
• Suno — AI music generation with vocal synthesis (freemium)
• MuseScore 4 — Free notation software with AI-assisted arrangement

2. Production

• Audacity — Free, open source audio editor. Noise removal, EQ, compression.
• Ardour — Free open source DAW (digital audio workstation). Full multitrack recording.
• LMMS — Free, open source beat-making and music production.

3. Sound Design

• AI can generate sound effects from description
• Use spectral analysis tools to clean up AI-generated audio
• Layer AI outputs with real recordings for authenticity

4. Mastering

• LANDR — AI mastering (freemium)
• Audacity can handle basic mastering chains for free
• A/B compare against reference tracks in your genre

Tips for AI Music

• AI excels at generating starting points — melodies, chord progressions, beats
• It struggles with precise arrangement and emotional nuances
• Best workflow: AI generates raw material → you arrange, edit, and mix
• Always check licensing terms on AI-generated audio before publishing

Video & Motion

Workflow: Prompt → Video

1. Script and plan

• Use AI to generate scripts, shot lists, and storyboards
• "Write a 60-second script about [topic] for YouTube. Conversational tone, include B-roll suggestions."

2. Create

• DaVinci Resolve — Free professional video editor with AI features (magic mask, speed warp)
• Blender — Free 3D animation, VFX, and motion graphics
• CapCut — Free editor with AI captions, background removal
• Runway — AI video generation and editing (freemium)

3. Enhance

• AI-powered upscaling for older footage
• Auto-captioning (DaVinci Resolve and CapCut do this natively)
• Background removal and replacement
• Color grading with AI-assisted scene matching

4. Distribute

• Render multiple aspect ratios (16:9, 9:16, 1:1) for different platforms
• AI can generate thumbnails and title variations for A/B testing

Writing & Documentation

Workflow: Prompt → Published Content

1. Research and outline

"Create a detailed outline for a technical blog post about [topic]. Include section headings, key points per section, and suggested code examples."

2. Draft

• Use AI for first drafts, then edit heavily — your voice matters
• Best for: technical documentation, README files, tutorials, API docs
• Worst for: personal essays, opinion pieces (sounds generic)

3. Edit and refine

• LanguageTool — Free grammar and style checker (open source)
• Vale — Free prose linter, configurable style rules
• Hemingway Editor — Free readability checker

Knowledge Management

This is where most builders fail. You generate incredible amounts of content and research — keeping it organized is the real skill.

Tools (All Free):

Best Practices:

• One source of truth — pick one notes app, use it for everything
• Link, don't duplicate — reference other notes instead of copying content
• Daily notes — capture fleeting ideas, process them weekly
• Project folders — each project gets a folder with README, decisions, and reference material
• Version control your notes — Obsidian vault in a Git repo = full history, free backup

Data Management for Builders

Regardless of medium, you're generating data: files, versions, prompts, outputs, configurations. Here's how to keep it manageable:

Storage Strategy

~/projects/
  project-name/
    README.md          # What this project is, how to build/run
    DECISIONS.md       # Why you made key choices
    prompts/           # Prompts that produced good results
    src/               # Source files (code, CAD, project files)
    output/            # Generated artifacts
    assets/            # Reference material, images, docs
    .git/              # Version control

Backup Rules

1. Code and text — Git + remote (GitHub, GitLab, Codeberg — all free)
2. Large files — Git LFS for binaries under 1GB, cloud storage for larger
3. 3D models and STLs — Version with Git LFS or store in Printables/Thingiverse
4. Audio and video — External drive + one cloud backup. Too large for Git.
5. AI model weights — Don't store locally unless you need to. Re-download is easier.

Prompt Archiving

Your prompts are intellectual property. Save the good ones:

## Prompt: Isometric Room Generator
- **Model:** Stable Diffusion XL + ComfyUI
- **Date:** 2026-03-15
- **Prompt:** "isometric cozy room, warm lighting, plants, bookshelf, pixel art style, detailed, no people"
- **Negative:** "blurry, low quality, text, watermark"
- **Settings:** Steps 30, CFG 7, Seed 4821937
- **Result:** output/isometric-room-v4.png
- **Notes:** Best at 1024x1024. Lower CFG gives more creative results.

Putting It All Together

The most powerful projects combine mediums:

• App + Hardware: Design a sensor enclosure in OpenSCAD, print it, connect to a Raspberry Pi running your Python code
• Music + Video: AI-generate background music, sync it to a video essay edited in DaVinci Resolve
• Art + Code: Generate textures in ComfyUI, use them in a Three.js web experience
• Writing + Design: Draft documentation in Obsidian, generate diagrams with Mermaid, publish with Astro or Hugo

The tools don't matter as much as the workflow. Find a loop that works — describe, generate, refine, ship — and run it fast.