Daniel Beglaryan

FROSTPROMPT

by FrostNode

|

Intelligent prompt engineering toolkit for Nano Banana Pro and Nano Banana 2. Semantic search, quality scoring, batch generation, multi-model formatting, and iterative editing chains.

0+Lines
0Modules
0Commands
0Prompts
0Models
ZeroDependencies

Capabilities

What FrostPrompt searches, scores, and generates

Keyword Search

Search 11,795 curated prompts by keyword with scored ranking, category filtering, and style matching.

Semantic Search

AI-powered search using sentence embeddings. Finds conceptually similar prompts even without shared keywords.

Quality Analyzer

Scores any prompt A+ to F against 10 criteria from Google's official guidelines. Returns specific improvement suggestions.

Auto-Tagger

Automatically tags all 11,795 prompts across 4 dimensions: style (20 tags), use-case (19 tags), subject (12 tags), and mood (10 tags).

Prompt Templates

7 built-in reusable templates with variable substitution — quote cards, product shots, portraits, landscapes, and more.

Batch Generation

Generate multiple prompt variations at once. Vary lighting, camera, color grading, style, or composition independently.

Prompt Chaining

13 iterative editing operations for refining images step by step — change background, lighting, style, text, outfit, season, and more.

Multi-Model Formatting

Adapt any prompt for 6 different models: Nano Banana Pro, Nano Banana 2, Midjourney, DALL-E 3, Flux, and Stable Diffusion.

Export System

Export prompts as JSON, Markdown, plain text, or ready-to-use Gemini API payloads. Save to file or clipboard.

History & Favorites

Track every prompt you generate, save favorites with tags, search your history, and re-use past work.

Five Google Frameworks

Built on Google's official 5 prompting frameworks: Generation, Editing, Real-Time Info, Text Rendering, and Creative Director mode.

Zero Dependencies

Core toolkit runs on Python 3.8+ with zero external dependencies. Semantic search optionally adds sentence-transformers.

Use Cases

Real-world scenarios where FrostPrompt accelerates your creative workflow

Social Media Content Creator

Create scroll-stopping visuals for Instagram, TikTok, and YouTube. Search existing prompts by style, fill templates with your brand text, and batch-generate variations for A/B testing.

$ python src/templates.py fill social-post -v subject "artisan coffee flat-lay" -v text_overlay "BREW BETTER"
social-mediatemplatesbatch

E-Commerce Product Photography

Generate professional product photography prompts with precise lighting, camera specs, and material descriptions. Adapt the same prompt across multiple aspect ratios.

$ python src/batch.py ratios "A minimalist ceramic jar on marble" --ratios 1:1 4:5 16:9
ecommerceproduct-shotmulti-ratio

Brand & Marketing Team

Maintain visual consistency across campaigns. Use templates for brand guidelines, chain edits to iterate on hero images, and export ready-to-use Gemini API payloads.

$ python src/export.py single "your prompt" --format api --model gemini-2.0-flash --ratio 16:9
marketingexportapi-integration

Multi-Platform Creator

Write one prompt and adapt it for every image model you use. FrostPrompt formats for Nano Banana, Midjourney, DALL-E 3, Flux, and Stable Diffusion with model-specific optimizations.

$ python src/multimodel.py all "A cinematic portrait in golden hour light"
multi-modelcross-platformadaptation

Prompt Engineering Learner

Analyze what makes prompts effective. The quality analyzer scores against 10 criteria and tells you exactly what's missing — lighting, camera, composition, materiality, and more.

$ python src/analyze.py "A photo of a sunset"
educationanalysisbest-practices

Content at Scale

Generate dozens of prompt variations from a single concept. Batch across subjects, lighting setups, color grades, and styles. Chain follow-up edits for iterative refinement.

$ python src/batch.py vary "A portrait in a studio" --count 10 --dimensions lighting camera color
batchscalevariations

Prompt Engineering Dashboard

Search, analyze, and generate — all from one interface

FrostPrompt v1.0.0
cyberpunk portrait11,795 prompts
KeywordSemantic
Style: CyberpunkMood: DramaticSubject: Portrait
Cyberpunk Neon Portrait97%

A hyperrealistic cyberpunk portrait bathed in neon pink and cyan light...

portraitcyberpunkneon
Futuristic City Street94%

A rain-soaked Tokyo alley at midnight with holographic advertisements...

landscapesci-fimoody
Neon Samurai Warrior91%

A samurai warrior standing in a cyberpunk marketplace with glowing katana...

charactercyberpunkdramatic
Quality AnalyzerA
Subject Clarity
Lighting
Camera Angle
Color Palette
Composition
Materiality
Atmosphere
Style Reference
Negative Space
Text Rendering

Supported Models & Platforms

One prompt, every model — with platform-specific optimizations

Nano Banana Pro

Full Support

Nano Banana 2

Full Support

Midjourney v6

Prompt Adaptation

DALL-E 3

Prompt Adaptation

Flux

Prompt Adaptation

Stable Diffusion XL

Adaptation + Negatives

Platforms

WindowsmacOSLinux

Export Formats

JSONStructured data, pipelines
MarkdownDocumentation, sharing
Plain TextClipboard, direct paste
Gemini APIDirect API integration

Quick Start

Zero dependencies — clone and run

Clone & Run (Core — Zero Dependencies)
1> git clone https://github.com/dbeglaryan/FrostPrompt.git
2> cd FrostPrompt
3> python src/search.py search "product photography" --limit 5
4> python src/analyze.py "your prompt here"
5> python src/templates.py fill quote-card -v quote_text "Hello World"
Enable Semantic Search (Optional)
1> pip install sentence-transformers
2> python src/semantic.py build # One-time: builds embeddings cache (~4 min)
3> python src/semantic.py search "peaceful nature scene" --limit 5
Batch Generation
1> python src/batch.py vary "A sunset over mountains" --count 5 --dimensions lighting color style
2> python src/batch.py subjects --template "Product shot of {subject}" --subjects "sneakers" "watch" "perfume"
Export as Gemini API Payload
1> python src/export.py single "your prompt" --format api --model gemini-2.0-flash --ratio 9:16

Pipeline Architecture

Six-phase prompt engineering pipeline

1

Input

User describes what they want — a concept, rough idea, or existing prompt to improve.

2

Search & Match

Keyword and semantic search across 11,795 prompts. Auto-tagging filters by style, use-case, subject, and mood.

3

Framework Selection

Selects the right Google framework — Generation, Editing, Real-Time, Text Rendering, or Creative Director.

4

Enhance & Score

Applies best practices, analyzes against 10 quality criteria, iterates until Grade A+.

5

Adapt & Batch

Formats for target model(s), generates variations, creates editing chains for refinement.

6

Export & Save

Outputs as JSON, Markdown, plain text, or API payload. Saves to history and favorites.

How FrostPrompt Compares

Feature comparison with popular alternatives

FeatureFrostPromptawesome-nano-banana-pro-promptsGeneric Prompt Generators
Prompt Database11,795 curated11,795 (static list)0-100
Semantic SearchYesNoNo
Quality Scoring (A+ to F)YesNoNo
Prompt Templates7 built-inNoLimited
Batch GenerationYes (5 dimensions)NoNo
Prompt Chaining13 edit operationsNoNo
Multi-Model Support6 modelsNano Banana only1-2 models
Auto-Tagging (4 dimensions)Yes (61 tags)NoNo
API Payload ExportYesNoNo
History & FavoritesYesNoNo
Google's 5 FrameworksYesNoNo
External Dependencies0 (core)Node.js + CMSVaries
CLI Commands350Varies