GRANDI-AI

Boundless Ideas · Limitless Reconstruction

GRANDI-AI professional AI image creation platform
Build extraordinary visual worlds, recover exceptional detail, and push creative images further.

Enter Studio
SYS.01
ANALYSIS
DEPTH:MAX
MODE:ON
DEEP ANALYSIS
ANALYSIS MODE
Challenge Analysis

Why can’t perfect image enhancement be achieved in one click?

Image super-resolution is not simple pixel stacking. Every stage—from diverse sources to competing goals—introduces uncertainty.

01

Extreme source diversity

Photography, AI art, hand-drawn originals, matte-painting composites, and CG renders all have different texture and damage patterns. Each needs its own treatment.

Photography AI art Hand-drawn Matte painting CG render
02

Competing creative goals

Resolution, structural repair, character resemblance, and creative detail often pull in different directions and must be balanced carefully.

Resolution Structure repair Character resemblance Creative detail
01 / 02

There is no universal one-click solution—but there can be an engine built to push the limits.

GRANDI-AI ECOSYSTEM

A multi-model enhancement engine

GRANDI-AI is a deeply optimized multi-model ecosystem. Rebuilt processing cores and tuned parameter coupling provide a flexible image-restoration workflow.

LOGIC // 01

Interactive enhancement

Work beyond automation limits with an iterative strategy: base upscale — local redraw — second enhancement.

EDITING // 02

Full local editing

Intervene precisely in damaged regions with local remove / generate / edit / optimize operations.

DIVIDE // 03

Divide and conquer

For dense or severely damaged images, crop key regions, enhance them independently, and merge the finished pieces.

GRANDI-AI gives you the tools to turn severely damaged, detail-starved images into
production-quality realistic artwork.

PX-01
100 MPTiled upscale + iterative enhancement + image editing
Structure:2
Creativity:1/10
Resemblance:4/3
Fractal:-2
Result image
Original image
Original
Result
NEW-01
8KImage edit + upscale
Structure:0
Creativity:-4
Resemblance:4
Fractal:-2
Result image
Original image
Original
Result
6KInk style + fidelity restoration
Structure:0
Creativity:1
Resemblance:0
Fractal:-1
Result image
Original image
Original
Result
Z-90
8KPowerful repair
Structure:2
Creativity:10
Resemblance:3
Fractal:-2
Result image
Original image
Original
Result
A-02
8KPowerful repair
Structure:2
Creativity:10
Resemblance:3
Fractal:-2
Result image
Original image
Original
Result
V-RT
8KStandard enhancement ×2
Structure:2
Creativity:1
Resemblance:3
Fractal:0
Result image
Original image
Original
Result
4KSharpen + denoise
Structure:1
Creativity:10
Resemblance:3
Fractal:-2
Result image
Original image
Original
Result
5KStandard enhancement ×2
Structure:0
Creativity:1
Resemblance:1
Fractal:0
Result image
Original image
Original
Result
4KDeblur
Structure:0
Creativity:0
Resemblance:2
Fractal:0
Result image
Original image
Original
Result
TX-07
4KStandard enhancement
Structure:0
Creativity:2
Resemblance:2
Fractal:0
Result image
Original image
Original
Result
4KStandard enhancement
Structure:0
Creativity:0
Resemblance:0
Fractal:0
Result image
Original image
Original
Result

Frequently asked questions

FREQUENTLY ASKED QUESTIONS
Q.01
How does GRANDI-AI billing work?

GRANDI-AI uses memberships and credits. Credits included with a membership expire with that plan. Separately purchased credits are not a membership benefit and remain valid for two years. Redeemed credits are non-refundable.

> Image upscale: 1 start credit + 1 credit per megapixel (about 5 credits for 4 MP; 33 credits for 32 MP).
> Structure repair or text-poster mode: total cost × 1.2.
> Detail Refinement: total cost × 1.1; with Structure Repair the multipliers stack to 1.32, then credits are rounded up once.
> Local protection: +1 credit per region, or +2 with structure repair.

> Image editing and generation: model-dependent.
> Color match: 1 start credit + 1 credit per image.
> Image cutout: 2 credits per task.
> Prompt enhancement: 2 credits per request.

Unlimited members can enable Unlimited Mode. When enabled, tasks can be submitted without spending credits (GPT-Image2 and NanoBanana2 & Pro models are excluded). Unlimited Mode tasks enter the queue at medium priority. During busy periods (usually weekday afternoons), high-priority credit tasks move ahead of them. After being passed for up to 20 minutes, a task is automatically upgraded to high priority. A queued no-credit task can also be manually upgraded to a credit task to raise its queue priority. Turn off Unlimited Mode to submit new tasks as credit tasks by default.

Standard membership is priced around ¥0.10 for 1.4 credits; Professional, quarterly, and annual plans have better rates, while Basic has a higher rate.

Q.02
How does GRANDI-AI protect data and privacy?

We process account information, uploads, instructions, results, history, and necessary logs only to provide tasks, display results, bill credits, protect the service, diagnose issues, and support users. Reasonable safeguards include encrypted transmission and access controls. Uploads and generated results are not used for unauthorized purposes such as AI model training.

When you delete a task card, images with no remaining task references are automatically cleaned up within 24 hours.

If you choose a third-party API model (the NanoBanana series or GPT-Image2), required materials and instructions are sent to that provider. Assess that path before uploading private, confidential, unpublished, or otherwise sensitive content.

Report account anomalies, suspected leaks, infringement, or unlawful content to contact@grandiai.com or WeChat cganimitta.

Q.03
What types of images can GRANDI-AI upscale?

GRANDI-AI is optimized primarily for realistic imagery, with limited support for stylized 3D, ink painting, oil painting, and related styles.

It is particularly strong at creative upscaling and detail enhancement for hand-drawn work, composites, 3D renders, and AI-generated storyboards. Product-faithful upscaling and old-photo restoration are supported on a limited basis.

Q.04
Does upgrading remove watermarks from earlier images?

No. Watermark status is locked when a task is submitted.

Free-user trial tasks remain watermarked in display and download, even after a later upgrade.

Conversely, tasks submitted during a paid membership remain watermark-free and downloadable after the membership expires.

After upgrading, use “Apply parameters” on the earlier task card, keep the random seed, and submit a new task to obtain a watermark-free result.

Q.05
How can I improve resemblance?
  • Lower Creativity to -3 or below.
  • Raise Resemblance to 4 or above.
  • Lower Fractal to -2 or below.
  • Use a higher output resolution (at least 16 MP).
  • Protect small, high-frequency details with a Local protection mask.
Q.06
What should I do if the upscaled image contains heavy noise or unwanted artifacts?
  • Switch Optimization mode to Sharpen & denoise.
  • Lower Fractal to -2 or below.
  • Avoid excessive Creativity (recommended: ≤ 2).
  • Add a short negative instruction, such as “clean wall without cracks or weathering; face without freckles.”