A: Hey Username3!

Yes, absolutely! AgenticFlow is well-suited to automate the creation and (to an extent) posting of YouTube videos on a periodic basis from prompts.

Here’s a general outline of how you could set that up:

Content/Script Generation (from Prompts):
Input: Your "certain prompts" would be the starting point. This could be a topic, a keyword, a link to an article to summarize, etc.
LLM Node: Use an LLM node (our built-in economical models or your BYOK) to expand on your prompt and generate a video script. You can instruct it on tone, length, style, and even suggest visual ideas.
Example Prompt for LLM: "Create a 2-minute video script about 'The Future of Renewable Energy,' including an intro, 3 key points, and an outro. Suggest simple visual concepts for each section."

Voiceover Generation:
Text-to-Speech Node: Take the generated script and feed it into our text_to_speech node (which offers various voices) or the text_to_speech_voice_clone node if you have a specific voice you want to use. This will output an audio file (MP3).
Visual Asset Generation/Selection:
AI Images: Use our generate_image node. You can use segments of your script or the visual concepts suggested by the LLM in step 1 as prompts to create relevant still images.

AI Video Clips (Short): For more dynamic visuals, you could use our text_to_video or image_to_video_v3 nodes to generate short AI video clips based on parts of your script or the generated images.
Stock Media (via API Call): If you use a stock media service with an API (like Pexels, Storyblocks, etc., if they have an API accessible via our API Call node), you could have a step that searches for and downloads relevant b-roll footage based on keywords from your script.

Video Assembly:
Render Video Node: Use our render_video node. This node is like a simple video editor controlled by instructions. You'll provide it with:
The voiceover audio file from step 2.
The URLs of your generated images or video clips from step 3.
Instructions on timing (how long each visual should appear), transitions (if supported by the node or future enhancements), and any text overlays (like titles or captions from the text_to_speech node's caption file).
Reference Template: Our "VIDEO - STORY THROUGH EMOJI" (https://agenticflow.ai/app/explore/workflow/40b8c32a-cdd4-45fd-b369-5864915f2ac4) is a good example of how different elements are combined into a video. You'd adapt it by using your script/prompts instead of emojis.
Posting to YouTube (Periodic Basis):
YouTube Data API MCP: Connect your YouTube account using our YouTube Data API MCP (https://agenticflow.ai/mcp/youtube_data_api).
Upload Video Action: The MCP has an action to upload videos. You'd feed the video file generated by the render_video node into this action, along with the title, description, tags, and privacy status.
Scheduling:
Current Method: To run this entire workflow "on a periodic basis" (e.g., daily, weekly), you'd currently use an external scheduler (like a cron job or a service like Pipedream/Make) to call your AgenticFlow workflow's API endpoint at the desired interval.
Roadmap: A native, built-in scheduler node within AgenticFlow is on our roadmap (please vote for it here: https://agenticflow.featurebase.app/p/scheduler-node-to-trigger-workflow-on-interval).

In Short:
Yes, you can create a workflow in AgenticFlow to:
- Generate a script from your prompts (LLM).
- Create a voiceover (Text-to-Speech).
Generate/select visuals (Image/Video Gen nodes or API Calls).
Assemble these into a video (Render Video).
- Upload to YouTube (YouTube MCP).
Schedule it to run periodically (API trigger for now).

This can become quite a sophisticated content engine! Start by building each piece, then chain them together.
— Sean