1. Puter Account — Sign-in and Limits

The "☁ Puter Account" block manages your connection to the AI provider.

• Guest Access: Start using the tool immediately without registration via a temporary guest session. Note that progress and accumulated credits are not saved between visits.
• Registration: The "🪪 Sign in / Register" button opens the Puter.com login/signup window. Registered accounts receive their own monthly AI credit quota, which persists across browser sessions.
• 🔄 Refresh account: Manually refreshes your login status and remaining credit balance.
• Credit Indicator: Displays monthly AI credit usage (in percentages and absolute values), remaining credits, and total monthly limit. The progress bar changes color dynamically: Green (normal usage) → Yellow (70%+ used) → Red (90%+ used).
• 📊 Usage Statistics: Detailed breakdown of processed images and consumed credits per model for the current month, along with an estimated "Remaining Photos" metric per model.
• ℹ️ Model Comparison: Comparison table showing estimated monthly image processing capacity under Puter's standard free tier, alongside quality ratings (Excellent / Good / Basic / Experimental). Estimates are calculated based on benchmark tests with 10 sample images per model and may vary depending on image complexity and generated metadata length.

2. Settings — Generation Options

2.1. AI Model

Dropdown menu of available models (Google Gemini, Alibaba Qwen, xAI Grok, Meta Llama). Models differ in credit cost and processing capabilities — Puter credit usage per image varies across models. The recommended default model is gemini-2.5-flash-lite, offering the best balance of credit efficiency and metadata generation quality based on testing results, allowing you to process the highest volume of images on Puter's free quota.

2.2. Platform

Selecting a target platform determines specialized prompt templates tailored to stock agency requirements and default character/keyword limits (customizable manually). Includes 7 platform-specific profiles and 6 universal profiles. Click "🧭 Platform Guide" to open an in-app reference overview of requirements and features directly in the interface without opening agency documentation.

2.3. Title & Description Max Chars

Numeric input fields specifying maximum character lengths. Toggle switch behavior:

• Enabled: The field is generated by the AI within the defined character limit.
• Disabled: The field is omitted from AI requests, saving credits. The output remains blank for manual editing or for platforms that do not require it (e.g., Vecteezy no longer requires a separate description).

When switching platforms, character limits and toggle states adjust automatically based on the selected profile, but can always be adjusted manually — changes are saved in the browser.

2.4. Keywords Count

Specifies how many keywords to request from the model. Defaults are set according to agency guidelines, but can be freely customized to any value.

2.5. Top Keywords Count

Defines how many of the initial keywords in the list are considered most critical for agency search algorithms. This is passed into the prompt as an instruction for the model: rank the first N keywords by importance (Primary Object → Primary Action → Secondary Details → Environment → Category → General Concepts). These same first N keywords are visually badged with a rank number in the keyword chip editor so you can instantly see which words carry the most weight.

2.6. 🔑 Keyword Rules — Personal Overrides

Global rules applied on top of selected platform profiles:

• Include Keywords: Custom terms that must always be present in the final keyword list. Inserted immediately after the "Top N" position to prevent accidental removal by limits or deduplication.
• Exclude Keywords (Stop-list): Personal blacklist: terms automatically removed from results, regardless of what the model generated.
• Trademarks: List of brand names (e.g., "iPhone", "Nike", "Coca-Cola") to automatically strip from keywords. Pre-filled with popular brands and editable to fit your niche.
• Keyword Format: Options include Auto (platform default), Single words only, or Allow 2-word phrases.
• Auto-apply Rules: Automatically strips blacklisted terms, trademarks, and inserts forced keywords immediately after each generation when enabled.
• Auto-remove Duplicates: Strips duplicate terms (case-insensitive) automatically right after generation.

3. Universal Profiles

General-purpose profiles for personal archives, scenarios with similar requirements across platforms, or asset types that don't fit specialized profiles:

4. Specialized Platform Profiles

Profiles tuned to match official microstock guidelines (rules verified as of mid-2026):

5. Image Upload and Generation

5.1. Uploading Assets

Drag & Drop files into the dropzone or click to open file picker. Supports JPEG, PNG, WebP, GIF, TIFF, BMP up to 100 images per session. Unsupported files are automatically rejected with a notification explaining why.

5.2. Processing Workflow

Clicking "✨ Generate AI Metadata" initiates queue processing (sequentially) with adaptive rate-limit throttling: if the provider rate-limits, pauses increase; during stable operation, pauses decrease. Progress counter shows progress ("N / M").

For each image:

1. Resized locally in browser memory (without server upload) to optimal dimensions for the selected model — saves credits without detail loss.
2. Submitted to the AI model alongside platform-specific prompt instructions.
3. Response is parsed and post-processed: platform stop-lists applied, personal Include/Exclude/Trademarks applied, duplicates removed, and orientation tags calculated.
4. Summary banner below Keywords field displays automated adjustments (words removed as stop-list/trademarks, include words added, duplicates cleared).

On errors (e.g., rate limits), requests retry automatically up to 3 times with progressively longer pauses between attempts. Individual cards can be re-generated manually via "✨ Generate again".

6. Metadata Card Interface

6.1. Title & Description

Auto-resizing text fields equipped with live character counters (Green = within limit, Orange = exceeded) and one-click clipboard copy buttons.

6.2. Interactive Keyword Chip Editor

Keywords rendered as visual chips:

• Add / Delete: Type and press Enter or comma (or paste comma-separated lists). Click × or press Backspace on empty input to delete.
• Inline Edit: Double-click any chip to edit text directly.
• Reorder: Drag-and-drop chips via mouse.
• Visual Highlights: Top N rank badging, yellow outline (duplicate warning), red outline (stop-list or trademark alert), green outline (Include keyword).

6.3. Individual Card Tools

Dropdown menu above the Keywords field with targeted operations applied to the current card:

• 🏷️ Insert include keywords — manually insert (or top up) missing terms from Include Keywords.
• 🔁 Remove duplicates — strip repeated keywords.
• ✂ Split into words — break multi-word phrases into single words.
• 📏 Trim to limit — cut the list down to a set keyword count.
• 🚫 Remove trademarks — strip terms from the trademark list.
• 🛑 Remove stop-list words — strip terms from your personal stop-list.

After each operation, a short result appears below the field (e.g., "Removed: 3 duplicates" or "No trademarks found").

6.4. Global History (Undo/Redo)

All keyword adjustments — via the chip editor, individual card tools, or Batch Tools — are saved in a single page-wide history. Ctrl+Z undoes the last action, Ctrl+Y (or Ctrl+Shift+Z) redoes it. Batch edits applied across multiple cards undo simultaneously in one click. A brief toast at the bottom of the screen confirms what was just undone or redone.

6.5. 💾 Save with EXIF

Saves the current image (JPEG only) with Title, Description, and Keywords embedded directly into EXIF/XMP fields — ready to verify in any image viewer or upload directly to a stock agency with the fields already filled in.

7. 🧰 Batch Tools — Bulk Processing

Slide-out side panel to execute single-click operations across all processed assets simultaneously: Insert include keywords, Remove duplicates, Split into words, Trim to limit, Remove trademarks, Remove stop-list words — the same set as the individual card tools (section 6.3), applied to the whole batch at once. After applying, a summary "Applied to N of M images" is shown, and the entire action can be undone with a single Ctrl+Z.

8. 📄 CSV Export — Bulk Table Export

Dedicated drawer for configuring and exporting results into a CSV file — a format accepted by most microstock agencies for batch metadata uploads.

8.1. Agency Presets

Each preset already includes the correct column set, order, and delimiter matching the agency's official upload template:

• 🌐 Universal (basic): Filename, Title, Description, Keywords.
• 🅐 Adobe Stock: Filename, Title, Keywords, Category, Releases.
• 🔴 Shutterstock: Filename, Description, Keywords, Categories, Illustration, Mature Content, Editorial.
• 🟢 Freepik (Magnific): Filename, Title, Keywords, Prompt, Model (uses ; as the delimiter, per the platform's format).
• 🟤 Dreamstime: Filename, Title, Description, Keywords, Category 1/2/3, Editorial, Free.
• 💛 Vecteezy: Filename, Title, Description, Keywords, License.
• 📷 Getty/iStock: same structure as Universal (the agency has no official bulk-upload CSV format; provided here for personal use).

8.2. Custom Field Builder

Beyond the ready-made presets, a full catalog of 20+ fields is available — each can be toggled on/off and freely combined in any order. Fields fall into four types:

• Auto-filled from generation: Filename, Title, Description, Keywords.
• Static text across the batch: Copyright, Creator, Email, Website, Country, License, AI Model, etc. — entered once and applied identically to every row in the export.
• Yes/No toggles across the batch: Editorial, Illustration, Mature Content, Free.
• Empty placeholders: Category, Prompt, and similar fields that require individual filling per image — left blank in the CSV, marked "will be inserted empty — fill in manually" to show it's an intentional omission, not an error.

Any manual change to the field set automatically switches the preset to "🛠 Custom" mode — the previously selected official preset is not lost and can always be restored from the dropdown.

8.3. Export Execution

The "📄 Export CSV" button compiles the file using the current values (including any edits made manually in the cards) for title/description/keywords, the correct column delimiter for the selected preset, and UTF-8 BOM encoding (ensuring clean Cyrillic rendering in Excel). The file downloads as stock_metadata_<preset>_<date_time>.csv.

9. 📦 Download All as ZIP

Packs all successfully processed assets into a single .zip archive: each image is saved with metadata already embedded into EXIF (the same mechanism used by the "💾 Save with EXIF" button on individual cards, section 6.5), ready for direct stock submission. A progress indicator shows the count of files packed.

10. Usage Tips & Recommendations

1. Start with the Recommended Model: Use gemini-2.5-flash-lite — it offers the best balance of quality and available generations on Puter's free quota. Try "heavier" models selectively if the default result doesn't satisfy you.
2. Configure Trademarks and Stop-list Once: Set these up for your shooting niche to avoid manually cleaning brands and unwanted words out of every result.
3. Use Include Keywords for Recurring Themes: Great for series/collection names that repeat across photos, so you don't have to type them manually for each image.
4. Keep Top Keywords Count Around 5–10: Most platforms weigh the first keywords heavily in search ranking; setting this value too high dilutes priority.
5. Disable Description for Platforms That Don't Need It: (e.g., Vecteezy) — this matches the platform's requirements and saves AI credits, since the field isn't requested from the model at all.
6. Always Verify Before Bulk Export: The chip editor and highlights (duplicates, stop-list, Top N) exist for quick visual checks without reading line-by-line.
7. Use Batch Tools for Repetitive Edits: If you notice after generating a batch that, say, all photos need duplicates removed, you don't need to open every card — one click in Batch Tools applies it to all of them, and it's fully undoable with Ctrl+Z.
8. Register a Puter Account Instead of Using Guest Mode: Guest sessions don't save progress or credits between visits, while a registered account has its own stable monthly quota.
9. Mind Each Platform's Keyword Requirements: For example, Dreamstime rejects submissions with fewer than 7 keywords, while Adobe requires exactly 49 — choosing the right platform profile (not Universal) up front saves time on manual adjustments later.
10. Save with EXIF vs CSV Export — Different Scenarios: EXIF is useful when the platform reads metadata directly from the file on upload; CSV is for platforms that require a separate tabular upload following an official template for large batches.