Viral Video Factory From Research

I want to create an AI agent that helps me generate and publish faceless YouTube and TikTok videos with high viral potential. First, I will provide links to existing YouTube channels. The agent should crawl those channels, analyze video performance (views, engagement, frequency), and extract the most successful topics, common patterns, and recurring keywords. Next, the agent should perform additional keyword research: Identify the top-performing keywords from those channels Use keyword tools and web research to find related keywords with high search volume and low competition Find trending and relevant video topics based on those keywords and online interest Then, for each validated topic, the agent should: Write a script optimized for short-form faceless video (45–60 seconds) Generate a voice-over using natural AI narration Select or generate relevant visuals (stock footage, animations, B-roll, or AI-generated scenes) that match the script Create a final video in vertical format suitable for YouTube Shorts and TikTok Additionally, there should be an option to: Generate one long-form video (3–10 minutes) for YouTube per topic, with longer scripts, voice-over, and extended visuals Output an SEO-optimized title and description using current YouTube algorithm best practices Generate 2–3 thumbnail variations The agent must allow me to input a list of topics and generate one video per topic. Once all videos are created, provide an option to automatically queue and upload videos to an existing YouTube channel multiple times per day. The final goal is to streamline the process of faceless content creation based on real-time research, with everything optimized for reach and visibility.

First, pre-entered info...

final-output2

• A series of MP4 files (short-form vertical videos, optionally one long-form horizontal video). • A small set of thumbnail images (e.g., JPG or PNG). • One text-based file (could be plain text or JSON) per video containing its SEO-optimized title and description. • (Optionally) A scheduling file (e.g., JSON or CSV) with timing and metadata for automated uploads.

subagents-extracted2

OVERVIEW OF SUBAGENTS: 1. Research & Keywords Subagent 2. Script & Voice Over Subagent 3. Video Assembly Subagent 4. Thumbnails & Metadata Subagent SUBAGENT 1: RESEARCH & KEYWORDS Name: "ResearchMaster" 1) #216 - Research Topic (Or Keyword) Deeply DESCRIPTION: Gathers insights about the given topics or competitor channels to identify trends and angles. INPUT NEEDED: Topic(s), YouTube channel URLs, or an overall niche. OUTPUT: A text summary (up to 2000 characters) of relevant findings. 2) #218 - Brainstorm Related Keywords DESCRIPTION: From the research summary, produce a short list of up to 5 closely related keywords, each with brief data points. INPUT NEEDED: The text output from #216. OUTPUT: A concise keyword list (up to 5 keywords). Final Output (Subagent 1): • A reference file (e.g., "topics_and_keywords.json") containing the summarized findings and up to 5 keywords. SUBAGENT REQUIRED INPUT: {youtube-channel-links}, {topic-list} SUBAGENT FINAL OUTPUT: [topics-and-keywords] SUBAGENT 2: SCRIPT & VOICE OVER Name: "ScriptVOGenerator" 1) #171 - Write Voice Over Script Based On Instructions DESCRIPTION: Creates one short-form script (45–60 seconds) and optionally one long-form script (3–10 minutes), guided by the research. INPUT NEEDED: Summarized research/keywords from Subagent 1, plus instructions (desired length, tone, etc.). OUTPUT: short_form_script.txt, and optionally long_form_script.txt. 2) #170 - Turn Script Into Voice Over MP3 DESCRIPTION: Converts the generated script(s) into a natural-sounding voice-over. INPUT NEEDED: The text script(s) from #171. OUTPUT: voice_over.mp3 (and optionally long_form_voiceover.mp3). 3) #223 - Summarize Multiple Files Into A Single Output DESCRIPTION: If multiple files are generated (e.g., two scripts and two MP3s), compile paths and references into one summary for easier handoff. INPUT NEEDED: All script text files and MP3 files from steps #171 and #170. OUTPUT: script_vo_summary.json (or similar), listing each file reference and key details. Final Output (Subagent 2): • script_vo_summary.json, referencing: – short_form_script.txt (and optionally long_form_script.txt) – voice_over.mp3 (and optionally long_form_voiceover.mp3) SUBAGENT REQUIRED INPUT: {topics-and-keywords}, {script-parameters} SUBAGENT FINAL OUTPUT: [script-and-voice-over] SUBAGENT 3: VIDEO ASSEMBLY Name: "VideoAssemblerAI" Short-Form (Vertical) Workflow: 1) #201 - Generate Vertical Text Slide Video From MP3 & Script DESCRIPTION: Creates a 9:16 text slide video synced with the short-form voice-over. INPUT NEEDED: voice_over.mp3 and short_form_script.txt. OUTPUT: base_short_form.mp4 (vertical). 2) #205 - Get Royalty-Free VERTICAL Videos For A Given Keyword (Optional) DESCRIPTION: Retrieves up to 2 relevant vertical stock footage clips based on a keyword. INPUT NEEDED: Keyword(s) from Subagent 1. OUTPUT: Up to 2 stock footage MP4 URLs. 3) #199 - Add Images & Videos On Top Of Existing MP4 DESCRIPTION: Overlays or splices additional visuals (stock clips or images) onto the base text slide video. INPUT NEEDED: base_short_form.mp4 and any stock/video assets. OUTPUT: final_short_form.mp4 Long-Form (Horizontal) Workflow (Optional): • #200 - Generate Text Slide Video From MP3 & Script (16:9). • #204 - Get Royalty-Free (Horizontal) Videos For A Given Keyword (up to 2 clips). • #199 - Combine footage and text slides. • OUTPUT: final_long_form.mp4 4) #223 - Summarize Multiple Files Into A Single Output DESCRIPTION: Gathers references to final_short_form.mp4 and optional final_long_form.mp4. INPUT NEEDED: final_short_form.mp4 and optionally final_long_form.mp4. OUTPUT: video_assembly_summary.json listing the paths of all MP4 outputs and any relevant details. Final Output (Subagent 3): • video_assembly_summary.json, referencing final_short_form.mp4 and optionally final_long_form.mp4. SUBAGENT REQUIRED INPUT: {topics-and-keywords}, {script-and-voice-over}, {assembly-parameters} SUBAGENT FINAL OUTPUT: [assembled-videos] SUBAGENT 4: THUMBNAILS & METADATA Name: "ThumbNMetaMaker" 1) #222 - Make Image (Especially With Text) DESCRIPTION: Creates up to 2 simple thumbnail images. INPUT NEEDED: Thumbnail text, brand style guidelines. OUTPUT: Thumbnail files (e.g., thumbnail1.png, thumbnail2.png). 2) #190 - Write or rewrite text based on instructions DESCRIPTION: Produces SEO-friendly titles and descriptions for each video; can also generate optional scheduling data. INPUT NEEDED: Topics, keywords, final video details, desired style. OUTPUT: "video_metadata.json" (or a text file) with titles, descriptions, and optional schedule. (Optional) #202 - Extract Thumbnail Images Of MP4 Video DESCRIPTION: Captures frames from the final video as possible thumbnails. INPUT NEEDED: final_short_form.mp4 or final_long_form.mp4 plus a prompt. OUTPUT: Up to 3 screenshot thumbnails. 3) #223 - Summarize Multiple Files Into A Single Output DESCRIPTION: Collect references to all thumbnails (1–2 generated images, optional screenshots) and the metadata file(s). INPUT NEEDED: The thumbnails from #222 or #202, plus video_metadata.json. OUTPUT: thumbs_and_metadata_summary.json listing all thumbnail paths, the metadata file location, and optional schedule info. Final Output (Subagent 4): • thumbs_and_metadata_summary.json, referencing: – 1–2 (or up to 3, if screenshots used) thumbnail images – video_metadata.json (titles, descriptions) – (Optional) scheduling_file.json/csv for upload times SUBAGENT REQUIRED INPUT: {assembled-videos}, {topics-and-keywords}, {metadata-parameters} SUBAGENT FINAL OUTPUT: [thumbnails-and-meta]

subagentnumber

[subagentnumber]

Then, agent response...

subagent1refined-new-json-core

I'll now add section G) TASKFLOW SUMMARY to Subagent 1's documentation, extracting and detailing each task from the workflow: G) TASKFLOW SUMMARY Task 1: - ID: #216 - Title: Research Topic (Or Keyword) Deeply - Description: Analyzes provided YouTube channels and topics to extract performance metrics, trends, and patterns. Includes view counts, engagement rates, posting frequency, and content themes. - Input(s): {youtube-channel-links}, {topic-list} - Output: [research_summary] Task 2: - ID: #218 - Title: Brainstorm Related Keywords - Description: Processes research findings to identify and evaluate relevant keywords. Generates 4-10 keyword suggestions with accompanying performance metrics and potential viability. - Input(s): [research_summary] - Output: [keyword_ideas] Task 3: - ID: #223 - Title: Summarize Multiple Files Into A Single Output - Description: Combines research summary and keyword list into a structured JSON format. Filters to top 5 keywords and formats all data according to specified structure. - Input(s): [research_summary], [keyword_ideas] - Output: [topics_and_keywords] Final Output Structure: ```json { "summary": "string (up to 2000 chars)", "keywords": [ { "keyword": "string", "data_points": "string" } // ... up to 5 keyword objects ] } ``` This taskflow represents the complete sequence of operations needed for Subagent 1 to transform raw channel/topic inputs into a structured research and keyword analysis output.

subagent1mermaid

<!DOCTYPE html> <html> <head> <title>Mermaid Diagram Example</title> <script src="https://cdnjs.cloudflare.com/ajax/libs/mermaid/9.4.3/mermaid.min.js"></script> </head> <body> <div class="mermaid"> graph TD A[Start SubAgent: ResearchMaster] --> B[Input Collection] B --> C[Channel URLs] B --> D[Topic List] C --> E[Task #216: Research Topic Deeply] D --> E E --> F[research_summary.txt] F --> G[Task #218: Brainstorm Keywords] G --> H[keyword_ideas.txt] F --> I[Task #223: Create Final JSON] H --> I I --> J[topics_and_keywords.json] J --> K[End SubAgent] subgraph Silo 1: Research C D E F end subgraph Silo 2: Keywords G H end subgraph Silo 3: Assembly I J end </div> <script> mermaid.initialize({ startOnLoad: true }); </script> </body> </html>

Mermaid Image - https://static.aiz.ac/1746309742-mermaid/mermaid1.png

prompts-required

[prompts-required]

subagent1-json-array

[ { "taskID": 216, "semanticTitleOfTask": "Research Topic (Or Keyword) Deeply", "taskDescription": "Analyzes YouTube links and topic(s) to identify trending content, engagement metrics, and patterns for successful videos, then summarizes key findings in a concise format (up to 2000 characters).", "inputDescription": "Requires one or more YouTube channel URLs plus a list of topics to investigate. Gathers info on channel performance, audience engagement rates, posting frequency, and emerging content trends.", "inputRequired": [ "youtube-channel-links", "topic-list" ], "outputDescription": "A single text output containing a structured research summary (up to 2000 characters) that highlights performance insights about the provided channels/topics.", "outputName": "research_summary", "promptInstruction": "As a YouTube content research specialist, analyze the provided channel(s) and topic(s). Focus on content patterns, typical view counts, engagement rates, posting frequencies, top-performing videos, and current audience trends. Compile a concise summary of findings (max 2000 chars) emphasizing practical insights for upcoming content." }, { "taskID": 218, "semanticTitleOfTask": "Brainstorm Related Keywords", "taskDescription": "Processes the research summary to find 4–10 relevant, high-potential keywords or phrases, noting search volume, competition, and potential for audience interest in short-form content.", "inputDescription": "Takes the research summary from Task #216 as its input. Uses that data to generate keywords with brief context on their viability, search volume, competition levels, and any notable trending behavior.", "inputRequired": [ "research_summary" ], "outputDescription": "A text file/list containing 4–10 potential keywords, each with brief performance notes or data points about search trends, relevance, and short-form video potential.", "outputName": "keyword_ideas", "promptInstruction": "As a keyword research specialist, examine the research summary to generate 4–10 promising keywords or phrases. For each, include a heading and 1–2 lines of data about search volume, competition, and trending patterns. Focus on keywords that align strongly with the original topic(s) and have high potential for short-form video success." }, { "taskID": 223, "semanticTitleOfTask": "Summarize Multiple Files Into A Single Output", "taskDescription": "Combines the research summary and the generated keyword list into a final JSON with top-level 'summary' and 'keywords' fields. Only the top 5 most promising keywords (with short data) should be included.", "inputDescription": "Needs the research summary from Task #216 and the keyword list from Task #218. This information is merged into a single JSON structure containing the summary and 5 selected keywords.", "inputRequired": [ "research_summary", "keyword_ideas" ], "outputDescription": "A final JSON object named 'topics_and_keywords' with two main fields: 'summary', containing the up to 2000-char research, and 'keywords', an array of up to 5 objects featuring 'keyword' and 'data_points'.", "outputName": "topics_and_keywords", "promptInstruction": "Combine the full research summary from Task #216 with up to 5 relevant keywords from Task #218 into a single valid JSON object. Use the field name 'summary' for the research text. For the selected keywords, include an array 'keywords' where each entry has 'keyword' and 'data_points'. Ensure proper JSON formatting and character escaping." } ]

subagent1-json-array-mermaid

<!DOCTYPE html> <html> <head> <title>Mermaid Diagram Example</title> <script src="https://cdnjs.cloudflare.com/ajax/libs/mermaid/9.4.3/mermaid.min.js"></script> </head> <body> <div class="mermaid"> graph TD A[Start ResearchMaster Subagent] --> B[Input Gathering] B --> C[Channel Links] & D[Topic List] C & D --> E[Task #216: Research Topic Deeply] E --> F[research_summary.txt] F --> G[Task #218: Brainstorm Keywords] G --> H[keyword_ideas.txt] F & H --> I[Task #223: Combine Into JSON] I --> J[topics_and_keywords.json] J --> K[End ResearchMaster Subagent] subgraph Silo 1: Research Stage C & D & E & F end subgraph Silo 2: Keyword Stage G & H end subgraph Silo 3: Assembly Stage I & J end </div> <script> mermaid.initialize({ startOnLoad: true }); </script> </body> </html>