Digital artists use Mage.space to iteratively polish, correct, and stylize visual art without starting over from a blank canvas or regenerating entire scenes. By pairing cloud-hosted Image-to-Image (Refine) transformations with surgical Inpainting (/apps/inpaint), creators take rough sketches, 3D viewport renders, or initial AI drafts and elevate them into production-ready character illustrations.
Instead of treating AI image generation as a one-click slot machine, professional illustrators and concept artists treat Mage.space as a high-speed layered painting assistant. This workflow eliminates local GPU hardware bottlenecks while providing granular control over anatomy, facial expressions, wardrobe details, and overall lighting cohesion.
1. The Core Split: Global Refinement vs. Localized Inpainting
A disciplined workflow separates global stylistic cohesion from localized structural corrections. Mage.space exposes both pathways across its web suite:
| Workflow Step | Mage.space Feature | Typical Use Cases | Key Advantage |
|---|---|---|---|
| Global Pass (Img2img) | Refine & Advanced Image Input | Turning sketches into rendered art, shifting lighting styles, upgrading texture fidelity across the whole frame | Locks core silhouette and camera composition while elevating rendering quality |
| Local Surgical Pass | Inpaint (/apps/inpaint) | Fixing anatomical flaws (hands, extra fingers, mismatched pupils), altering facial expressions, swapping props and garments | Modifies only masked pixels; preserves 95%+ of the approved artwork intact |
| Finishing Touches | Skin Enhancer & Image Enhancer | Restoring natural skin pores, clearing compression noise, upscaling to 2K or 4K | Produces clean, print- and display-ready output without altering character identity |
2. Image-to-Image Workflows: From Rough Concept to Finished Painting
Digital artists rarely feed empty prompts into Mage.space when working on client deliverables or personal graphic novels. Instead, they begin with visual scaffolding:
- Importing the Foundation Art: Artists upload an initial digital sketch, a rough grayscale color wash, a 3D blockout from Blender, or an earlier AI generation that had great composition but lacked finish.
- Setting Denoising and Guidance (Refine):
- Low Denoising (0.20–0.35): Preserves line art and color blocks strictly. It introduces subtle surface textures, metallic sheens, and fabric folds without changing character geometry.
- Medium Denoising (0.40–0.60): The standard "rendering pass." It transforms loose rough lines into finished digital paintings, semi-realistic character art, or anime cel-shading while retaining poses and framing.
- High Denoising (0.65+): Used only during visual brainstorming when artists want radical stylistic variations of an existing pose.
- Model and Concept Anchoring: On Mage.space, artists pair models like SDXL, Illustrious, Pony, or photorealistic checkpoints with curated LoRAs. This ensures the output adheres to specific art styles (such as cinematic concept art, dark fantasy oil, or modern webtoon) without drifting into generic aesthetics.
3. Inpainting Workflows: Surgical Precision for Character Refinement
Even the strongest generative models frequently output anatomical anomalies—such as warped fingers, mismatched iris colors, asymmetrical jewelry, or messy clothing seams. Mage.space Inpainting isolates and repairs these flaws without rolling back the rest of an approved piece.
Step 1: Applying the Mask with Soft Boundaries
In the Mage Inpaint tool (/apps/inpaint), the artist paints a mask strictly over the target area (for instance, a distorted hand or an unconvincing eye).
- The Margin Rule: Always extend the brush mask 10 to 15 pixels past the flawed area into clean anatomy and surrounding background. This buffer gives the diffusion model sufficient contextual context (skin tone, ambient rim light, fabric grain) to blend the repaired edge seamlessly into the untouched canvas.
Step 2: Isolated, Task-Specific Prompting
A common mistake is copying the full 40-word prompt used to generate the original image into the inpainting prompt box. In Mage Inpaint, the model concentrates entirely on the masked bounding box.
- Incorrect Inpaint Prompt:
masterpiece, 1girl, standing in rainy neon city, cyberpunk jacket, detailed face, perfect hands, 8k resolution - Correct Inpaint Prompt (Fixing Hands):
a slender hand resting naturally on the hip, five distinct fingers, realistic knuckles, soft skin tones, matching lighting - Correct Inpaint Prompt (Refining Eyes):
expressive detailed anime eyes, deep blue iris, sharp pupils, subtle catchlights, clean eyelashes
Step 3: Layered Iteration
Rather than trying to fix both hands, face angles, and costume armor in a single click, artists work sequentially:
- Fix facial expression and eyes first.
- Download or send the result directly back into Inpaint.
- Mask and correct hand anatomy or weapon grips.
- Mask and detail costume trim, belt buckles, or hair strands.
4. Advanced Character Consistency and Refinement Stack
For recurring characters across multi-panel comics or character concept sheets, artists combine Inpainting with Mage’s broader toolkit:
- Character Reference & Face Lock: Under higher tiers, artists feed a canonical portrait into the reference slot. When inpainting facial adjustments, this reference prevents the model from generating a completely different face.
- ControlNet Constraints: Available on Mage Pro, ControlNet lets artists lock the underlying pose skeleton (Openpose) or line contours (Canny/Depth), ensuring that adjustments to clothing or expressions never shift the original dynamic gesture.
- Skin Enhancer & Image Enhancer: Once anatomy and clothing details are locked, artists push the illustration through
Skin Enhancerfor balanced texture, followed byImage Enhancer(up to 2K on Pro, 4K on Pro Plus) to finalize pixel-level clarity.
5. Subscription Tiers & Commercial Value for Artists
Mage.space runs on a monthly subscription model with unlimited generation modes rather than a strict per-image token charge, making iterative inpainting cost-effective for heavy daily use:
| Tier | Monthly Price | Relevance to Digital Artists |
|---|---|---|
| Free | $0 | Basic exploration; queue speeds and access to advanced apps are constrained. |
| Basic | $10/month | Entry for standard refinishing: Unlimited base model generations, LoRAs and Textual Inversions, full access to Inpainting and Refine, and 500 monthly Gems. Ideal for freelance artists handling routine cleanups. |
| Pro | $30/month | Professional sweet spot: 4x generation speed, premium base models, unlimited character creation, ControlNet, Style & Face Reference, 2x LoRAs stackable, 2K Enhance, and 1,500 monthly Gems. Essential for serious character consistency. |
| Pro Plus | $60/month | Studio-grade power: Everything in Pro plus custom model imports (Checkpoints, LoRAs, Lycoris), 5x LoRAs, 4K Enhance, and 3,500 monthly Gems. Best for artists deploying bespoke trained styles. |
Practical Recommendation
For artists modifying existing line art, repairing AI imperfections, and refining character sheets, the Basic ($10/mo) plan provides sufficient access to the web-based Inpaint and Refine tools without burning through image credits. However, artists doing commercial character design who require Face Reference, ControlNet, and multi-LoRA styling will find the Pro ($30/mo) plan indispensable for repeatable production quality.