What Shall We Build Next?

Describe

Describe your task

Refine

Refine the plan

SubAgents

Review all agents

Deploy

Deploy your agent

A) SUBAGENT SUMMARY  
"ResearchMaster" gathers insights from provided YouTube channels and/or topic lists, then distills that research into a concise summary along with up to 5 high-potential related keywords.

B) FINAL TASK OUTPUT  
A single JSON file called “topics_and_keywords.json” containing:  
• A text summary of relevant findings (up to ~2000 characters).  
• A list of up to 5 closely related keywords (with brief notes or data points).

C) SUBAGENT INPUT  
{youtube-channel-links}, {topic-list}

E) SUBAGENT TASK SUMMARY

1) Input Gathering  
   - The subagent begins by taking the user-provided YouTube channel URLs and/or topic(s) directly from {youtube-channel-links} and {topic-list}.

2) Deep Research (#216 - Research Topic Or Keyword Deeply)  
   - Combines {youtube-channel-links} and {topic-list} as the “topic” to research.  
   - OUTPUT: research_summary.txt or a similar text string containing up to ~2000 characters of summarized findings.

3) Brainstorm Keywords (#218 - Brainstorm Related Keywords)  
   - Feeds the research_summary.txt from step 2 into #218.  
   - OUTPUT: A structured text list of 4–10 related keywords (with commentary or data).

4) Consolidate Into Final JSON (#223 - Powerful LLM Prompt-to-Text Response)  
   - Gathers (a) the research summary from #216’s output and (b) the keyword list from #218’s output.  
   - Instructs the LLM to format them into a final JSON object with two main fields: "summary" (the text from step 2) and "keywords" (up to 5 from step 3).  
   - OUTPUT: “topics_and_keywords.json”.

F) SILOS

SILO 1: Research Stage  
• INPUT(s): {youtube-channel-links}, {topic-list}  
• ACTION: #216 (Research deeply) → research_summary.txt

SILO 2: Keyword Ideation Stage  
• INPUT: research_summary.txt  
• ACTION: #218 (Brainstorm related keywords) → keyword_ideas.txt (4–10 keywords)

SILO 3: Final Assembly Stage  
• INPUT(s): research_summary.txt + keyword_ideas.txt  
• ACTION: #223 (Combine into a final JSON structure) → topics_and_keywords.json

This completes Subagent 1’s workflow, yielding a final “topics_and_keywords.json” file that captures the key research findings and up to 5 top keywords.

SubAgent #1 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY:  
"ScriptVOGenerator" takes research outputs and user parameters, then creates one short-form script (and optionally one long-form script) along with corresponding voice-over MP3 files. Finally, it bundles these files into a single JSON “summary” file for easy handoff.

B) FINAL TASK OUTPUT:  
• script_vo_summary.json (referencing):  
  – short_form_script.txt  
  – voice_over.mp3  
  – (optionally) long_form_script.txt  
  – (optionally) long_form_voiceover.mp3

C) SUBAGENT INPUT:  
• {topics-and-keywords} (the JSON or text file containing the summarized research and up to 5 keywords from Subagent 1)  
• {script-parameters} (instructions on desired script length, tone, style, and whether a long-form script is needed)

E) SUBAGENT TASK SUMMARY:  
1) #171 - Write Voice Over Script Based On Instructions  
   → INPUT: {topics-and-keywords} + {script-parameters}  
   → OUTPUT: short_form_script.txt (and optionally long_form_script.txt if requested)

2) #170 - Turn Script Into Voice Over MP3  
   → INPUT: The text script(s) generated in step #171  
   → OUTPUT: voice_over.mp3 (and optionally long_form_voiceover.mp3 if a long-form script was generated)

3) #223 - Powerful LLM Prompt-to-Text Response (Summary Step)  
   → INPUT: All generated files from steps #1 & #2 (e.g., short_form_script.txt, voice_over.mp3, [long_form_script.txt, long_form_voiceover.mp3])  
   → OUTPUT: script_vo_summary.json (lists file references, filenames, and any relevant metadata)

F) SILOS:

• SILO A (Short-Form Script & VO Generation)  
   1) #171 (Write short_form_script.txt)  
   2) #170 (Convert short_form_script.txt → voice_over.mp3)

• SILO B (Optional Long-Form Script & VO Generation)  
   3) #171 (Write long_form_script.txt)  
   4) #170 (Convert long_form_script.txt → long_form_voiceover.mp3)

• SILO C (Final Summarization)  
   5) #223 (Collect references to scripts + MP3s and output script_vo_summary.json)

SubAgent #2 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
This subagent, “VideoAssemblerAI,” takes in completed scripts and corresponding voice-over audio files (from Subagent 2), along with keyword information and assembly parameters, and produces one or more fully assembled MP4 videos (vertical short-form plus optional horizontal long-form). It also provides a summary JSON file referencing the final video files.

B) FINAL TASK OUTPUT  
The final output is a single JSON file (“video_assembly_summary.json”) that references the path/URL(s) to:  
• final_short_form.mp4 (9:16 vertical)  
• optionally final_long_form.mp4 (16:9 horizontal)

C) SUBAGENT INPUT  
• {topics-and-keywords} – The JSON or text output from Subagent 1 containing up to 5 keywords (needed if we want relevant stock footage).  
• {script-and-voice-over} – A JSON summary from Subagent 2 referencing “short_form_script.txt,” “voice_over.mp3,” and optionally “long_form_script.txt,” “long_form_voiceover.mp3.”  
• {assembly-parameters} – Instructions or toggles specifying whether to create only short-form, or also a long-form, plus any details regarding visuals (e.g., "use stock footage" or "use minimal overlays").

E) SUBAGENT TASK SUMMARY  
Below is the step-by-step flow of how VideoAssemblerAI will create the final video outputs:

1) Generate Base Short-Form Vertical Text Slide Video (#201)  
   • INPUT: “voice_over.mp3” + “short_form_script.txt.”  
   • OUTPUT: “base_short_form.mp4” (vertical 9:16).

2) (Optional) Retrieve Royalty-Free Vertical Stock Footage (#205)  
   • INPUT: One or more keyword(s) from {topics-and-keywords}.  
   • OUTPUT: Up to 5 relevant vertical MP4 URLs. (Typically, the workflow selects up to 2 to overlay/splice, ignoring the rest.)

3) Combine Additional Visuals with Base Short-Form Video (#199)  
   • INPUT: “base_short_form.mp4” plus up to 2 chosen stock footage URLs.  
   • OUTPUT: “final_short_form.mp4.”

(Optional Long-Form Workflow)  
4) Generate Base Long-Form Horizontal Text Slide Video (#200)  
   • INPUT: “long_form_voiceover.mp3” + “long_form_script.txt.”  
   • OUTPUT: “base_long_form.mp4” (horizontal 16:9).

5) (Optional) Retrieve Royalty-Free Horizontal Stock Footage (#204)  
   • INPUT: One or more keyword(s) from {topics-and-keywords}.  
   • OUTPUT: Up to 5 relevant horizontal MP4 URLs. (Typically, the workflow selects up to 2 to overlay/splice.)

6) Combine Additional Visuals with Base Long-Form Video (#199)  
   • INPUT: “base_long_form.mp4” plus up to 2 chosen stock footage URLs.  
   • OUTPUT: “final_long_form.mp4.”

7) Summarize All Outputs (#223)  
   • INPUT: “final_short_form.mp4” and optionally “final_long_form.mp4.”  
   • OUTPUT: “video_assembly_summary.json” – references final MP4 URLs, durations, or any relevant notes.

F) SILOS  
• Silo 1: Short-Form (Vertical)  
  1) #201 → “base_short_form.mp4”  
  2) #205 (Optional) → get vertical videos  
  3) #199 → “final_short_form.mp4”

• Silo 2: Long-Form (Horizontal) (Only if assembly-parameters specify a long-format output)  
  4) #200 → “base_long_form.mp4”  
  5) #204 (Optional) → get horizontal videos  
  6) #199 → “final_long_form.mp4”

• Final Step (Both Silos)  
  7) #223 → “video_assembly_summary.json”

This completes Subagent 3’s final, detailed workflow for assembling short-form and (optionally) long-form videos.

SubAgent #3 - Diagram

Expand Flow

A) SUBAGENT SUMMARY  
"ThumbNMetaMaker" creates one or more thumbnail images (optional additional screenshots from the final videos), then generates a text-based metadata file (title, description, optional schedule), and finally compiles references to these assets into a single JSON summary.

B) FINAL TASK OUTPUT  
• A JSON file named “thumbs_and_metadata_summary.json” referencing:  
  – 1–2 (or up to 3 optional) thumbnail image URLs  
  – “video_metadata.json” containing SEO-based titles and descriptions (and optionally scheduling data)

C) SUBAGENT INPUT  
• {assembled-videos}: Information and file paths to the final short-form and/or long-form MP4 files from Subagent 3.  
• {topics-and-keywords}: Core topics, keywords, or content research results to guide the thumbnail text and metadata.  
• {metadata-parameters}: Specific instructions on branding, style, or scheduling; and the details about how the user wants the titles/descriptions.

E) SUBAGENT TASK SUMMARY  
1) (Optional) #202 - Extract Thumbnail Images Of MP4 Video  
   DESCRIPTION: If user wants to automatically generate screenshots from the final video(s), this skill can provide up to 3 thumbnail images.  
   INPUT NEEDED: final_short_form.mp4 and/or final_long_form.mp4 from {assembled-videos}, plus a prompt (e.g., “Extract the best frames for eye-catching thumbnails”).  
   OUTPUT: Up to 3 thumbnail images (PNG or JPG) with short textual descriptions.

2) #222 - Make Image (Especially With Text)  
   DESCRIPTION: Creates 1–2 custom thumbnail images containing stylized text or brand elements.  
   INPUT NEEDED: Thumbnail text prompt or branding guidelines from {metadata-parameters}, possibly referencing {topics-and-keywords} (e.g., “Vertical travel thumbnail with big bold text ‘Top 5 Spots’ in a modern style”).  
   OUTPUT: 1–2 PNG files (e.g., “thumbnail1.png”, “thumbnail2.png”).

3) #190 - Write or rewrite text based on instructions  
   DESCRIPTION: Generates an SEO-friendly title and description for each video, referencing {topics-and-keywords} and {metadata-parameters}. May also include optional scheduling data.  
   INPUT NEEDED:  
      – Topics or keywords from {topics-and-keywords}.  
      – Final video details (length, format) from {assembled-videos} if relevant.  
      – {metadata-parameters} or user instructions defining the style/tone.  
   OUTPUT: "video_metadata.json" file with structured data  
      (example contents: title: “…”, description: “…”, tags: […], scheduleTime: “2023-11-01 12:00:00”, etc.).

4) #223 - Summarize Multiple Files Into A Single Output  
   DESCRIPTION: Collect all the generated thumbnail image URLs (from #202 and/or #222) and the “video_metadata.json” from #190. Produces a single JSON summary listing all assets.  
   INPUT NEEDED:  
      – List of thumbnail image files from steps #202 and/or #222  
      – "video_metadata.json" from step #190  
   OUTPUT: "thumbs_and_metadata_summary.json" referencing:  
      – All thumbnail image paths (1–2 custom images, plus any optional screenshots)  
      – The metadata file location (video_metadata.json)  
      – Optional scheduling data if present

F) SILOS  
• SILO 1: Thumbnail Generation  
   – (Optional) #202 to extract frames from the final MP4(s).  
   – #222 to generate 1–2 custom thumbnail designs with text.  
• SILO 2: Metadata Creation  
   – #190 to produce an SEO-focused “video_metadata.json.”  
• SILO 3: Aggregate Final Outputs  
   – #223 to unify all generated assets into “thumbs_and_metadata_summary.json.”