V1.50 Documentation

Introduction

NAI Smart Studio is a desktop app for working with NovelAI image generation. It combines Text-to-Image, Image-to-Image, and Inpaint with prompt editing, reference images, prompt search, and image history.

This manual covers the app’s controls. For parameters such as Steps, CFG, and Sampler, also see the official NovelAI documentation.

Screen layout

Left panel: Set Prompt / Negative, characters, generation parameters, continuous generation, and reference images.
Center panel: View and edit images in Canvas. Other tabs provide tag assistance, prompt search, metadata inspection, and more.
Right panel: View saved images in History. Click an image to display it in Canvas, or Shift+click to import its generation settings and other data.

Initial Setup

Get your API key

Log in to the NovelAI website.
Open account settings from the top-right account menu.
Get and copy your permanent API token.

Register the key and choose settings

In the app’s API Key Management area, paste the key into the Key field and click Add. Check the connection test result, then click Save in the settings dialog. If you register multiple keys, select one and click Activate to switch accounts.

Output folder: Choose where generated images are saved.
Wildcard folder: Choose where the app reads wildcard files.
UI language: Select Japanese, English, or another supported language.

Basic Generation

Enter what you want to draw and choose generation settings in the left panel, then click Generate. Results appear in Canvas and, once saved, can be viewed and reused from History.

Enter what to draw in Prompt and what to avoid in Negative.
Select a model and image size, then adjust other parameters as needed.
Check the estimated Anlas cost on the generation button and click Generate.

Prompt editing

Prompt / Negative: Enter tags or text. Use Tag Details to inspect tags and the token counter to check prompt length.
Characters: Use Add Character to set individual Prompt / Negative fields and positions. Disable a character to leave it out of generation; collapsing its fields does not disable it.
Find / Replace: Ctrl + F opens search and Ctrl + H opens replace across the main Prompt / Negative and all character fields.
Comments: Temporarily exclude text from generation. Ctrl + / toggles the selection or the item at the cursor; Ctrl + Shift + / or a line-number click toggles whole lines.

Basic parameters

Image Setting / Width / Height: Choose a square, portrait, or landscape preset, or set dimensions in 64px steps. Standard generation allows a long side up to 2048px and a short side up to 1536px.
Model: Select V3 / V4 / V4.5 / V5 (Full / Curated). V4 and later also offer Anime / Furry mode.
Steps / CFG / Sampler: Adjust the step count, prompt guidance, and sampling method.
Seed: -1 is random. A fixed value makes it easier to reuse the same conditions.

Advanced options include quality tags and CFG Rescale. V5 also offers light quality tags and transparent backgrounds. The estimated Anlas cost updates when you change the model, size, Steps, or other relevant settings.

Use Ctrl + Enter for normal generation or Ctrl + Alt + Enter to generate once. To repeat generation, see Continuous Generation.

Importing Images

Import saved images for editing or reuse their generation settings. Drag and drop an image, press Ctrl + O, Shift+click in History, or choose Import from an image’s right-click menu.

Choose a destination

Image-to-Image / Inpaint: Load the image into Canvas as a source image.
Vibe Transfer / Precise Reference: Add images to reference slots. You can import multiple images at once.
Tagger: Infer tags from the image.
NAI-info: Inspect the image’s embedded generation information.
Metadata: Select the fields to apply to the current input, such as Prompt, Negative, characters, generation settings, and Seed.

In the metadata options, enable Use unexpanded prompts to load the original wildcard or randomizer expressions. Leave it off when trying to reproduce the source image.

Image-to-Image / Inpaint

Generate from an existing image. Image-to-Image redraws the whole image; Inpaint redraws an area specified by a mask or selection.

Basic workflow

Load an image into Canvas and click i2i Mode below it.
To change only part of the image, draw a mask or select a Focus area. Without a selection, the app uses Image-to-Image.
Set the Prompt and strength, adjust noise or output size as needed, then click Generate.

Masks and image editing

Mask: Mark the area to change with Brush, Rectangle, or Polyline.
Focus: Temporarily enlarge a small area for Inpaint, then merge it back into the source image. The area inside the reference margin is changed. Clear the selection with the frame’s × or a right-click.
Paint: Draw directly on the image. Right-click to pick a color from it.
Edit: Crop, resize the canvas, move, scale, or rotate the image.

Image pending: images/I2I_EDIT_TOOLS_001_EN.png

Turn on Split Canvas to show the source on the left and generated results on the right. Use it to generate repeatedly while retaining the source image and mask.

Director Tools

Remove clutter or backgrounds, create line art or sketches, colorize, or change expressions. Use image history to return to the image before a transformation.

Image pending: images/I2I_DIRECTOR_TOOLS_001_EN.png

Normal I2I / Inpaint uses Width / Height for the output size. Focus keeps the source image’s dimensions. Use ← / → to move through image history. See Shortcuts for mask and editing undo controls.

Simple Upscale

Increase the resolution of a Canvas image or refine its details. Choose the custom or official method to suit your purpose.

How to use it

Load an image into Canvas and click Simple Upscale below it.
Select a method and adjust the scale and strength or level.
Check the output size and estimated Anlas cost, then run the upscale.

Choose a method

Custom Upscale: Requires Opus. Enlarges with RealESRGAN / Waifu2x, then regenerates in tiles. Choose an upscaler, scale (up to 4×), and strength.
Enhance (Official): Choose a scale supported by the model and source image and a level from 1–5. Max is V5-only. Even 1x regenerates the image; Advanced settings let you adjust strength and noise separately.

Anlas cost depends on the method, settings, and remaining allowance. Check the displayed notices and requirements. Stopping may wait for the active operation to return. Results are saved and appear in History.

Continuous Generation

Repeat generation with the same settings or work through candidates in the generation queue.

Open Continuous Generation in the left panel and choose a count, a duration, or a run that continues until stopped.
Set the delay between generations and what to do when an error occurs.
Start continuous generation. You can also use Ctrl + Shift + Enter.

If the generation queue contains items, those take priority over the current on-screen input.

Stopping may wait for the current request to finish. Authentication errors, insufficient Anlas, or rate limits stop the run; resolve the cause before restarting.

Vibe Transfer

Use reference images to guide the mood or style of a generation. You can combine multiple images in separate slots. Vibe Transfer is unavailable with V5.

How to use it

Open Vibe Transfer in the left panel and use + to add images or vibe files.
Adjust reference strength and information extraction, then enable the slots you want to use.
Generate. Slots that have not been encoded are processed first.

Save and reuse

Save and load .naiv4vibe / .naiv4vibeBundle files, or export an image with embedded Bundle data. Disable unused slots to keep them for later.

Precise Reference

Use reference images to guide character design or style. Precise Reference works with supported V4.5 models and is unavailable with V5.

How to use it

Open Precise Reference in the left panel and use + to add a reference image.
Choose whether to reference the Character, Style, or Both.
Adjust strength and fidelity, then generate.

You can register multiple images and enable or disable individual slots. Background preview visibility can be changed in settings.

NAI-info

Inspect the generation information embedded in an image. Review its Prompt, Negative, characters, Seed, and other settings, then reuse them for another generation.

Drop an image into NAI-info or send it from the import dialog.
Review the summary, prompts, generation settings, and raw metadata.
Use Send to text2img or Send to img2img to reuse the information.

New screenshot slot
Planned file: images/NAI_INFO_001_EN.png
Show the NAI-info tab with preview and generation information.

For images with Actual Prompt records, you can view and copy the resolved text. The option to load original expressions is shared with Importing Images.

Without embedded metadata, only available information such as image dimensions is shown.

Tagger

Infer tag candidates from images to help write a Prompt. Use Tagger to describe an image’s features or export tags for multiple images. Complete the initial setup through Tools or the setup settings first.

How to use it

Load an image into Tagger, or send it from the import dialog.
Select the model and General / Character / Copyright thresholds, then run tag inference.
Review candidates and scores, select the tags you need, and click Add selected tags to prompt. Use Add tags to prompt to add the results together.

Inspect tags and process folders

Tag information: View Japanese aliases and Prompt Helper categories. Use ? to open Danbooru Wiki.
Folder batch processing: Infer tags for multiple images and export them as caption text files.

Prompt Helper

Find tags through search or categories and add them to Prompt / Negative. Use Prompt Helper when you cannot recall a tag name or want to explore related expressions.

Find and add tags

Search by English tag, Japanese alias, or description, or browse categories such as body, clothing, and expressions.
Review tag details and related tags.
Show the destination Prompt / Negative tab, then double-click a tag or click Add to current tab.

New screenshot slot
Planned file: images/PROMPT_HELPER_001_EN.png
Show Prompt Helper search, categories, tag list, and detail area.

Favorites: Save frequently used tags for easy access.
Custom tags: Add your own tags and categories.

Story Prompt

Use NovelAI text generation to turn an outline into a story and image prompts for multiple scenes. Select a text generation model in the Story Prompt tab. Click Check Models to refresh the list of available models.

Image pending: images/STORY_PROMPT_001_EN.png

How to use it

Input: Set the outline, characters, scene count, rating, and Prompt Language, then generate the story.
Story: Review and edit the text. You can also regenerate individual scenes.
Cuts & Prompts: Use Break into Cuts to split the story into shots, then adjust each cut’s Prompt and image count.
Generate: Use Review Plan to select cuts and click Add to Queue. Check the image model and size in the main panel, then click Generate to start processing.

You can also specify fixed character tags and a Prompt preset. Save and reopen your work as a project, or export Generation YAML, Data YAML, or Data JSON.

Wildcards

Insert words or phrases from a set of candidates into Prompt. Use wildcards to generate variations in hair color, clothing, backgrounds, and more.

How to use them

Prepare TXT / YAML / JSON files in the configured Wildcard folder. For TXT, write one candidate per line.
Select a file or entry in the Wildcards tab, review its contents, and copy the expression.
Paste it into Prompt and generate. Use Dynamic Prompt for random choices or Sequential Prompt to use candidates in order.

Syntax

{A|B|C}: Pick one of the inline candidates at random.
__name__ / __folder/name__: Pick from a file.
__file/key__: Pick from a YAML / JSON key.
[SEQ:__name__]: Use candidates in order during continuous generation.

Switch to Edit to modify and save file contents, or use an external text editor. Blank and # lines in TXT files are ignored. After expansion, CHAR: lines become character prompts.

Templates

Save prompts and generation settings together, then restore them in one step. Use templates to switch between familiar setups or return to previous settings after another task.

What is saved

Prompts: Prompt / Negative, character-specific input, and character positions.
Generation settings: Basic parameters such as image size, model, Steps, CFG, Seed, and Sampler, plus advanced options such as quality tags.

Save and load

Prepare the Prompt and generation settings you want to save.
Use the save icon above Prompt to create or overwrite a template.
Select a template from the list above Prompt and apply it with the check button.

The Templates tab lets you inspect, apply, rename, and delete templates. Selecting an entry alone does not change the current input.

Applying a template replaces the current prompts and generation settings. Use Presets for tag additions and cleanup.

Presets

Save reusable tag additions, exclusions, and cleanup rules. Create presets for quality tags, styles, or other purposes and apply them to the current Prompt / Negative.

Available settings

Prefix / Suffix: Add common tags at the start or end of Prompt.
Common Negative: Add unwanted elements to Negative.
Auto-Exclusion: Remove specified words.
Cleanup: Remove duplicate tags and extra spaces.

Save and apply

Configure a preset in the Presets tab and save it with a name. Select it from the list above Prompt and apply it with the check button. Presets also apply during Text-to-Image generation; queued generation follows its Apply preset on generation setting.

Use Templates to save generation settings such as image size and model as well.

Prompt Search

Search prompts by tag conditions and add the results to the generation queue. Use it to try multiple prompt candidates in sequence.

Setup

For first use, run setup from Tools or Settings → Search. When setup finishes, enable Prompt Search to display its tab in the center panel. Tagger becomes available too.

Initial setup downloads a large dataset. Keep the app open until it finishes.

Search and generate

Search using included tags, excluded tags, and ratings.
Review the results, set the number to add, and click Add search results to generation queue.
Check the image model, size, and other settings, then click Generate to process the queue in order.

Additional controls

Output filtering: Exclude copyright, character, artist, clothing, feature, or color tags when adding results to the queue.
Prompt Upscaler: Enrich prompts using tag inference during queue processing.

Prompt Chunks

Save reusable prompt fragments and insert them as components when needed. Organize quality tags, styles, and common phrases into categories.

Setup

Enable Prompt Chunks in Settings and open the setup wizard. Authenticate with email / password or Google login assistance to use the Chunk tab.

Create and insert chunks

Create a category in the Chunk tab and add a chunk.
Enter its name and prompt content, then save.
Select the chunk and click Insert into Prompt, or double-click it to insert.

Use Refresh to retrieve the latest server-side content. While editing a Prompt, use Ctrl + P or @ to show chunk suggestions.

Tools

Set up helper features and enable or disable the tabs you want to use.

Image pending: images/TOOLS_001_EN.png

Hiding a tab keeps its settings, input, and installed data. Canvas and Tools always remain visible.

Settings & Other Features

Open File → Settings to configure output locations, input assistance, external integration, and other preferences.

Main settings

General: Manage API keys, UI language, output and favorites folders, and other general options.
Format / filename: Set the output format and file naming rules.
Assistance: Configure tag completion, Japanese aliases, highlighting, comma assistance, and related tags.
Search / Chunk: Manage optional features and check setup or authentication status.
Others: Configure logs and storage-related options.

Click a related tag to add it to Prompt; right-click to add it to Negative or hide it temporarily. Ctrl + Space focuses the first suggestion. Change candidate counts and categories in Assistance.

Image pending: images/RELATED_TAG_SUGGEST_001_EN.png

Eagle integration

If you manage images with Eagle, configure the destination and other options in Integration. Send generated images automatically or use an image’s right-click menu to send it manually, with prompts and generation information as tags and memos.

History and image actions

Folder selection: Changes both the displayed images and the destination for subsequent generated images.
Click / Shift+click: Display the image in Canvas / open the import dialog. You can also drag images to import them.
Heart: Copy the image to the favorites folder. Select a destination if one has not been set.
Select and delete: Ctrl+click selects multiple images. With History focused, Ctrl+A selects all and Delete moves the selected images to the Recycle Bin.
Right-click: Save, copy, send to Eagle, open the containing folder, or import.