Docs/Guide

Feature Overview

This page walks through Nephele's features along an illustrator's workflow. It isn't a full tutorial — it's a map: first see what problem each stage solves, then click into the matching doc for the actual steps.

If you just finished downloading, start with Getting Started. If you want to understand why Nephele doesn't include any drawing features, read the Design Philosophy.


The core idea: from "finding inspiration" to "shipping it"

Nephele's features hang off one real workflow:

01

Inspiration & References

Agent chatFind referencesLocal similar-image searchReference board3D preview
02

Create

Nephele stays out
03

Organize

Library browsingEagle compatibilityLibrary indexingFile search
04

Review / Refine

Creative reviewStyleDNAAI credential detection
05

Publish & Protect

Publish packagingMulti-platform uploadCreator dashboardDigital certificationEvidence collectionReverse image search
Stardial wheel — double-tap Ctrl or press a mouse side button to summon it anywhere

Each feature maps to a specific stage: finding references, organizing assets, reviewing your work, packaging for release, and securing your rights. The creation itself still belongs to you.


Where each feature lives in the app

Here's a reference table first, so you don't get lost once you're inside the app:

Where in the UIActual nameWhat it does
Left nav barWorkbenchTalk to Nephele; the Agent does the work for you
Left nav barToolboxEntry grid for all the tools (two pages)
Toolbox · Page 1Creative ReviewWork review + commercial assessment (grade / dimensions / price range)
Toolbox · Page 1Digital CertificationIssue RFC 3161 timestamps, generate forensic reports
Toolbox · Page 1Publish PackagingMulti-spec image output + watermarking + steganographic protection
Toolbox · Page 1Evidence CollectionLock down infringement evidence, generate tamper-proof evidence packages
Toolbox · Page 2Reverse Image SearchTrace where an image came from
Bottom of windowLibraryBrowse local files and your Eagle library
Toolbox · Page 2Library IndexingBatch-tag your library with PixAI Tagger
Toolbox · Page 2StyleDNA5-axis style analysis, generates an artist identity card
Toolbox · Page 2Multi-Platform UploadAuto-generate post copy and tags for 7 platforms
Toolbox · Page 2Color PickerPick a color from anywhere on screen
Left nav barCreator DashboardAggregate publishing data across platforms
Double-tap CtrlStardial Quick WheelRadial quick menu to launch apps / switch views / open platforms
Agent chatReference BoardPin reference images to the screen and build a moodboard
Drag a file / AgentAI Credential DetectionDetect AI-generation markers
Double-click inside Eagle3D Model PreviewQuick PMX model viewer
Left nav barSettingsAccount, network, proxy, local models
Left nav barAboutVersion info, changelog

技巧

Your frequently used tools show up automatically in the "Recently Used" area of the left nav bar, so you don't have to dig through the Toolbox every time.


Global shortcut: the Stardial Quick Wheel

At any time, double-tap Ctrl (you can rebind it in Settings) and a two-ring radial menu pops up in the center of the screen. This is Nephele's Stardial — no clicking, no window switching; just hold Ctrl and move the mouse to reach any feature.

The first ring has 8 categories, laid out clockwise along the illustrator's workflow:

DirectionCategoryDefault items
CanvasPhotoshop, CSP, Blender, SAI
ReferenceEagle, Library Indexing, Color Picker, Reverse Search, Summon Reference Board
PublishPublish Packaging, Multi-Platform Upload
RightsDigital Certification, Evidence Collection, AI Credential Detection
PlatformsPixiv, ArtStation, Twitter, Xiaohongshu, Weibo
AnalysisCreative Review, StyleDNA
SystemFile Explorer, Screenshot tool, Terminal
NepheleMain window, Quick input, Toolbox, Settings, Edit Stardial

There are two ways to summon it: double-tap Ctrl, or press the mouse side button (fourth button / XButton1). The latter suits a mouse or stylus better: hold the side button to pop it up, move the mouse to select, release to execute — no keyboard at all.

Right-click the center of the wheel to open the Stardial Editor, where you can add or remove categories and items yourself. It supports launching any app, jumping to a Nephele view, opening a URL, or running a system command.

技巧

Tablet users can map their pen's side button to the fourth mouse button in the driver, so a pen-button press summons the Stardial — smoother than reaching for Ctrl.


Inspiration & References

Talking to Nephele (Agent)

The home page is the chat interface. You can just say things like:

  • "Find me some cyberpunk-style reference images"
  • "Check out the trending illustrations on Pixiv lately"
  • "Save the references I just picked into Eagle"
  • "Find a few images in Eagle with a similar style to this one"

Nephele will plan the search, call the right sources, and present the results in an image picker. You can preview, multi-select, save locally, import into Eagle, or pin straight to the reference board.

技巧

For detailed Agent usage, see Talking to Nephele.

Reference search sources

For detailed reference-search usage, see the Finding References guide.

Nephele's reference search falls into two categories:

  • Out of the box: Pixiv / Konachan cloud image libraries, Wikimedia Commons (real-world / encyclopedic image sources), and your local Eagle library
  • Requires the Wisp browser bridge: Pinterest / ArtStation / Huaban, collected through your own real browser

It lays out the search results as cards; you can tick the ones you want and save them locally or to Eagle in one click.

Local similar-image search

Upload an image and Nephele finds visually similar references in your Eagle library. It uses the EVA02 tag-embedding vectors produced as a byproduct of tagging for cosine similarity (85% weight), with tag overlap as a fallback ranking signal (15% weight). When you're hunting for references, your own library often understands you better than the internet does.

Reference board (pinning images)

The most annoying part of gathering references is switching windows back and forth. Nephele's reference board lets you pin all the images you need into a single always-on-top window and arrange them freely, like building a moodboard:

  • Drag to move, scroll to zoom, rotate via handle/toolbar
  • Cram in plenty of images — positions auto-stagger to avoid overlap
  • Lock mode: press Ctrl+Shift+L to lock, and the window becomes click-through, so you can work on the Photoshop or CSP underneath right through it
  • Unlock with the same shortcut
  • The Agent can pin found references straight onto it in one click

The reference board is borderless, takes up no taskbar slot, and when locked its background is fully transparent — only the images show, so it never blocks your canvas.

For detailed usage of the reference board and 3D preview, see Reference Board & 3D Preview.

3D Model Preview

Open a PMX model file in Nephele and a standalone 3D preview window pops up. It supports mouse rotate/zoom/pan, a three-viewport mode, and lighting plus IBL ambient-light adjustment. You can quickly check a model's pose and structure without opening Blender or MMD — very handy when working with 3D references.


Create

This is your territory, and Nephele stays completely out of it.

We don't provide brushes, layers, or any drawing features at all. In this stage you use Photoshop, CSP, Blender, or whatever tool you like, and Nephele won't get in your way.

What Nephele does is stand beside you and take care of the chores that have nothing to do with creating — finding references, organizing assets, writing copy, packaging output — so you can give that saved time back to your brush.


Organize: the library

Library

The file browser at the bottom of the window — move your mouse to the bottom-center of the screen to summon it. It unifies local folders and your Eagle library into a single browsing interface:

  • Local files — quick access to Desktop / Documents / Pictures / Downloads, or pin your frequently used project folders
  • Eagle library — browse the folder tree with expand/collapse, search, and sorting
  • Right-click actions — pin as reference, send to pipeline, copy path
  • Lightbox — double-click for a large preview, arrow keys to switch images

You can dig through assets without switching windows, then pin what you find to the reference board or drag it into the pipeline.

Library Indexing

Batch-process your whole library in one click:

  • PixAI Tagger content tags — automatically recognizes what's in each image and writes the Chinese display names straight back to your Eagle items; local ONNX inference, zero cloud cost
  • Image-search index — produces EVA02 tag-embedding vectors as a byproduct of tagging, used to rank local similar-image search
  • Semantic search index — gte-multilingual-base multilingual embeddings, so you can search your library across languages, including in Chinese
  • Resumable — if it crashes halfway, the next run picks up from where it left off without re-spending effort

Files that have already been tagged are marked Nephele-tagged and won't be processed again.

Let the Agent organize your library

Beyond batch tagging, you can also just tell Nephele:

  • "Import all the illustration-type images from my Downloads folder into Eagle and tag them by style"
  • "Move the images tagged 'mechanical' in my 'reference' folder over to the 'mecha reference' folder"
  • "Add a 'to review' tag to this batch of images"
  • "Empty the Eagle trash"

Nephele will search, classify, tag, and move folders on its own — you just confirm. For repetitive chores like organizing assets, handing it to the Agent is far faster than clicking around manually.


Review / Refine: let AI be your advisor

Creative Review

Not sure about a finished piece? Hand it to Nephele, and it dynamically adjusts the assessment dimensions based on the type of work:

General dimensions (60%): composition, color, creativity, finish, tonality

Specialized dimensions (usually 40%; for UI design and Logo/branding it flips to 60%, varying by type):

  • Illustration → technique, lighting, form, sense of space, detail
  • Chibi → form appeal, deformation consistency, expressiveness
  • UI design → information hierarchy, usability, visual consistency, adaptability
  • Logo/branding → recognizability, scalability, brand fit, simplicity
  • Pixel art → same dimensions as illustration (technique, lighting, form, space, detail)
  • Photo retouching → same dimensions as illustration (technique, lighting, form, space, detail)

Its main mode now is Creative Review: what you pulled off, where you could push further, what to try on the next piece — and it maps out a learning path for directions worth investing in long-term. When you want numbers, switch to the Commercial Assessment view — a ten-tier grade of S, S-, A+, A, B+, B, C+, C, D, E, a dimension radar, plus a market price-range reference. It's not about making you worship a score, but about offering an "outside observer's perspective."

For detailed assessment dimensions and the full workflow, see the Creative Review guide.

StyleDNA (StyleDNA)

MBTI for art style. Five bipolar axes analyze your drawing style:

AxisLeft endRight end
Line vs MassLP
Realistic vs StylizedRS
Minimal vs ComplexMX
Vivid vs MutedVU
Narrative vs MoodNA

32 combinations generate your personal artist identity card, which you can screenshot and share. Note: this isn't "instructing" you how to draw — it helps you understand yourself after you've finished, like an athlete reviewing game footage.

For a full breakdown of the 5-axis bipolar system, see the StyleDNA guide.

AI Credential Detection (AI metadata detection)

If you take commissions or enter competitions and need to confirm whether an incoming image is AI-generated, this feature scans EXIF, IPTC/XMP, C2PA credentials, and generation-parameter fingerprints. It does credential and metadata analysis — it doesn't use AI to "look at" the image content. It can currently recognize signatures from the following tools:

  • Midjourney
  • ComfyUI
  • Stable Diffusion (including SD.Next / Forge and A1111 parameter fingerprints)
  • NovelAI
  • InvokeAI
  • Fooocus
  • Gemini (Google)
  • DALL-E
  • Leonardo.ai
  • Adobe Firefly
  • Bing Image Creator

OpenAI images are recognized via C2PA content credentials rather than name matching.

It can also parse AIGC markers under China's "Labeling Method for AI-Generated Synthetic Content" (GB 45438-2025), recognizing generation markers written by domestic models such as Doubao, Jimeng, Tongyi, Wenxin, and Kling.

There are four possible result states: AI credentials found, weak signals, insufficient credentials, and no credentials found. Note that "insufficient credentials / no credentials found" only means no AI markers were detected — it does not confirm the work is human-made.

注意

Metadata can be modified or deleted, so detection results are for reference only. This isn't courtroom evidence — it's just the first filter.

For the full list of supported tool signatures plus C2PA parsing, see the AI Credential Detection guide.


Publish: from packaging to copy to rights

Publish Packaging

Output watermarked high-resolution PNGs in one click. It supports full-image watermarks (centered or tiled) and steganographic protection (a blind watermark, 256-bit embedding, with no visual impact). PSD project files can be processed directly, with output going to the same folder as the source file or a folder you specify.

For detailed parameters and how the blind watermark works, see the Publish Packaging guide.

Multi-Platform Upload

Find writing social-media copy a headache? Upload your work and Nephele auto-generates copy and tags tailored to 7 platforms:

  • Twitter (X)
  • Weibo
  • Xiaohongshu
  • Pixiv
  • ArtStation
  • Bilibili
  • Douyin

Tag formatting, copy tone, and character limits are all adapted per platform. The results show up right in the interface — just copy and post.

提示

On the desktop, Multi-Platform Upload handles generating the copy, tags, and publish-ready content; actually filling that content into a browser draft goes through the Wisp browser bridge. Nephele will never publish on your behalf without your confirmation.

For platform-rule details and copy templates, see the Multi-Platform Upload guide.

Creator Dashboard

After publishing, the real hassle is reviewing the numbers. The Creator Dashboard pulls the view counts, favorites, and engagement data from the platforms you're logged into back to your local machine, making it easy to review which platforms, subjects, and posting times suit you best.

The data channel relies on the Wisp browser bridge: it reads data from the creator backends you're already logged into and lands it in your own local database.

Digital Certification

A complete rights-securing workflow:

  1. Timestamp issuance — compliant with the RFC 3161 international standard, backed by professional authorities such as DigiCert
  2. Smart grading — when a source project file such as a PSD is detected, it's automatically upgraded to "full certification"; finished images alone go through "lite certification"
  3. Merkle tree hashing — in batch certification, all files are chained together, so tampering with any one breaks the entire tree
  4. PDF forensic report — a professional certificate with a QR code, ready to submit directly to a platform or court
  5. .nep certification package — original files + timestamp + report bundled together, with a complete chain of evidence

For the forensic report layout and .nep package structure, see the Digital Certification guide. You can verify .nep files online at verify.arisfusion.com.

Evidence Collection

Spotted infringement? Don't just take a screenshot. Nephele will help you:

  • Auto-capture the entire web page (a scrolling full-length screenshot)
  • Save the page source
  • Extract all the images on the page
  • Grab the TLS certificate info
  • Apply RFC 3161 timestamps to everything

It generates a tamper-proof evidence package you can use later when consulting a lawyer or filing a platform complaint.

For the full evidence-collection workflow and legal advice, see the Evidence Collection guide.

Reverse Image Search (origin tracing)

Upload an image and several engines search its origin in parallel at once:

  • Local library — searches your own Eagle library first, fully local and offline
  • MD5 exact match — hits the very same original image across major image sites by file-byte hash, with zero false positives (only a 32-character hash is sent)
  • SauceNAO — the strongest for tracing anime-style images
  • IqDB — joint search across multiple image sites
  • Google Vision — general image recognition
  • Baidu Image Search — coverage of the Chinese internet

Handy for tracking down stolen art and verifying where a reference came from.


Eagle library integration

If you manage your assets with Eagle, Nephele goes well beyond just "being able to read it":

  • Browse, search, and drag assets from Eagle into Nephele
  • Import references straight into Eagle (with tags / ratings / notes)
  • Update the tags, ratings, classification, and folders of Eagle assets through Agent chat

Nephele achieves all this by reading and writing Eagle's underlying data files directly — no plugin or API involved. More importantly: Nephele's operation and asset management don't depend on Eagle at all. For a detailed explanation of the relationship, see Eagle Compatibility & Independence.

Last updated Jun 21, 2026·Applies to v0.5.2-beta