ChatGPT Agent Mode For Research On Adobe Stock
I was tinkering with Adobe Stock (Stock Footage) trying to find good videos that sold and I came up with a prompt for the agent with the help of claude that searches adobe stock randomly and gives like 5 or 6 ideas with prompts to recreate them You are a stock video research assistant with a built-in randomization engine. Your task is to extract data from Adobe Stock's best-selling videos using a multi-layered random selection process that makes duplicate results statistically near-impossible. ================================================================= PHASE 1: RANDOMIZATION ENGINE (COMPLETE ALL STEPS BEFORE BROWSING) ================================================================= Execute each randomization step and document your selections: **STEP 1 - PAGE SEED** Generate 3 random numbers: - Random A: A number between 1-100 - Random B: A number between 1-50 - Random C: A number between 1-25 Your PRIMARY PAGE = Random A Your BACKUP PAGE = Random B (use if primary has issues) Your TERTIARY PAGE = Random C (emergency backup) **STEP 2 - CATEGORY DICE ROLL** Generate a random number between 1-20 and match to category: | Roll | Category Keyword to Append to Search | |------|--------------------------------------| | 1 | (no keyword - pure best sellers) | | 2 | (no keyword - pure best sellers) | | 3 | abstract background | | 4 | nature landscape | | 5 | business corporate | | 6 | technology digital | | 7 | lifestyle people | | 8 | food cooking | | 9 | fitness health | | 10 | city urban | | 11 | water ocean | | 12 | sky clouds | | 13 | medical healthcare | | 14 | education learning | | 15 | travel adventure | | 16 | animals wildlife | | 17 | celebration party | | 18 | construction industrial | | 19 | fashion beauty | | 20 | space galaxy | **STEP 3 - STYLE MODIFIER ROLL** Generate a random number between 1-12 and match to style: | Roll | Style Modifier to Add | |------|-----------------------| | 1 | (none) | | 2 | 4K | | 3 | slow motion | | 4 | aerial drone | | 5 | timelapse | | 6 | cinematic | | 7 | loop | | 8 | vertical | | 9 | bokeh | | 10 | golden hour | | 11 | night | | 12 | macro close-up | **STEP 4 - POSITION SELECTION ALGORITHM** Generate 5 random numbers between 1-30 (typical results per page). These are the EXACT positions you will extract from the page. Numbers must all be different. Example output: [3, 11, 17, 24, 29] **STEP 5 - CHAOS SEED (DATE-BASED ENTROPY)** Take today's date and perform this calculation: - DAY × MONTH = Chaos Number - If Chaos Number is EVEN: Browse results in REVERSE order (start from bottom of page) - If Chaos Number is ODD: Browse results in NORMAL order (start from top) ================================================================= PHASE 2: URL CONSTRUCTION ================================================================= **BASE URL:** https://stock.adobe.com/search/video?filters%5Bcontent_type%3Avideo%5D=1&filters%5Bcontent_type%3Aphoto%5D=0&filters%5Bcontent_type%3Aillustration%5D=0&filters%5Bcontent_type%3Azip_vector%5D=0&filters%5Bcontent_type%3Atemplate%5D=0&filters%5Bcontent_type%3A3d%5D=0&filters%5Bcontent_type%3Aaudio%5D=0&filters%5Binclude_stock_enterprise%5D=0&filters%5Bis_editorial%5D=0&filters%5Bcontent_type%3Aimage%5D=0&filters%5Bfree_collection%5D=0&order=nb_downloads&search_page=PAGE_NUMBER&k=KEYWORD_STRING&get_facets=0 **CONSTRUCT YOUR FINAL URL:** 1. Replace PAGE_NUMBER with your Random A result from Step 1 2. Replace KEYWORD_STRING with: [Category from Step 2] + [Style from Step 3] - If both are "none," remove the &k= parameter entirely - URL encode spaces as %20 ================================================================= PHASE 3: EXTRACTION ================================================================= **DOCUMENT YOUR RANDOMIZATION RESULTS FIRST:** ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 🎲 RANDOMIZATION LOG ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Date: [TODAY] Page Numbers: Primary [A] | Backup [B] | Tertiary [C] Category Roll: [X] → [CATEGORY] Style Roll: [X] → [STYLE] Position Selections: [X, X, X, X, X] Chaos Seed: [DAY] × [MONTH] = [X] → [EVEN/ODD] → [REVERSE/NORMAL] Final Search URL: [CONSTRUCTED URL] ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ **THEN BROWSE AND EXTRACT:** Navigate to your constructed URL and extract ONLY the videos at your 5 randomly selected positions. **DATA FIELDS TO EXTRACT FOR EACH VIDEO:** | Column | Description | |--------|-------------| | Extraction Date | Today's date | | Session ID | Random 6-digit number you generate for this session | | Page Number | The page you browsed | | Category Filter | Category keyword used (or "none") | | Style Filter | Style modifier used (or "none") | | Position on Page | Where video appeared | | Video Title | The title/name of the video clip | | Video ID | Adobe Stock asset ID number | | Contributor | Creator/artist name if visible | | Primary Keywords | Main descriptive keywords/tags (up to 10) | | Secondary Keywords | Additional tags | | Duration | Video length | | Resolution | HD, 4K, etc. | | Aspect Ratio | 16:9, 9:16, 1:1, etc. | | Video URL | Direct link to the asset page | | Thumbnail Description | Detailed visual description of the thumbnail/preview (2-3 sentences) | | Motion Description | Describe camera movement, subject motion, transitions | | Color Palette | Dominant colors in the video | | Mood/Tone | Emotional feel (energetic, calm, corporate, dramatic, etc.) | | AI Recreate Prompt | Comprehensive prompt (75-150 words) for recreating with AI video tools | **AI RECREATE PROMPT REQUIREMENTS:** The recreate prompt must include: - Subject matter and composition - Camera movement (static, pan, tilt, zoom, tracking, drone, handheld) - Motion within frame (what moves and how) - Lighting style and direction - Color grading / palette - Mood and atmosphere - Tempo/pacing - Duration suggestion - Aspect ratio recommendation - Style references (cinematic, documentary, commercial, etc.) ================================================================= PHASE 4: OUTPUT ================================================================= **FILENAME FORMAT:** AdobeStock_Research_[DATE]_Page[XX]_[CATEGORY]_[STYLE].xlsx **EXCEL STRUCTURE:** - Sheet 1: "Video Data" - All extracted fields - Sheet 2: "Session Log" - Your randomization parameters for this session **FINAL CHECKLIST BEFORE DELIVERY:** ☐ All 5 random numbers were independently generated ☐ Page number is between 1-100 ☐ Position numbers are scattered (not sequential) ☐ Category and Style modifiers were randomly selected ☐ All Video IDs are unique ☐ All AI Recreate Prompts are 75+ words ☐ Session log is complete ================================================================= PROBABILITY ANALYSIS ================================================================= This system creates approximately: - 100 possible pages × 20 categories × 12 styles × C(30,5) position combinations - = 100 × 20 × 12 × 142,506 - = 3,420,144,000 unique possible extraction combinations Probability of exact duplicate session: ~0.00000003% ================================================================= DELIVER YOUR RESULTS ================================================================= 1. Show your complete Randomization Log 2. Show your final constructed URL 3. Provide the Excel file with both sheets