Automated AI YouTube Shorts Pipeline

I want to build a web-based AI YouTube Shorts automation platform that can automatically create, schedule, and publish Shorts to multiple YouTube channels.

The main goal is to make the process as automated and low-cost as possible. The system should use FREE and OPEN-SOURCE AI tools wherever possible instead of relying on expensive paid APIs. The architecture should also allow AI providers to be changed later.

The platform should have a simple dashboard where I can connect multiple YouTube channels and select the correct channel for each video.

For example:

Channel: Sad Stories
Content Type: AI Story
Topic: Sad Dog Story
Duration: 30–45 seconds
Publish Time: 8:00 PM

When I click “Create & Schedule”, the system should automatically generate an original story based on the topic, create a script and hook, divide it into scenes, generate suitable visuals, create AI narration, add subtitles/captions, combine everything into a vertical 9:16 video, and schedule it to the selected YouTube channel.

The system should automatically generate the video title, description, hashtags and other basic YouTube information.

I also need a second automation for AI video clipping. I should be able to paste a YouTube video URL (only for content I own or have permission to reuse). The system should analyze/transcribe the video, identify interesting sections using AI, automatically create multiple Shorts, crop them to 9:16, add captions and hooks, and allow me to schedule them to a selected YouTube channel.

The platform should support multiple channels, with each channel securely connected through YouTube OAuth/API. It is very important that the video is always uploaded to the correct channel selected by the user.

The system should include a content queue showing the channel, topic, scheduled time, generation status, rendering status, publishing status, YouTube URL and errors. Failed jobs should be retryable.

For automation, I should be able to set schedules such as 3 Shorts per day at specific times. The system should automatically generate and publish content according to the queue.

Please prioritize FREE/open-source technologies such as Ollama/local LLMs, ComfyUI, Stable Diffusion/SDXL, Whisper/faster-whisper, Piper TTS, FFmpeg, Python, Laravel and MySQL/SQLite wherever practical.

Do not make the application dependent on one expensive AI provider. AI services should be configurable so I can replace them later.

The system should be mobile-friendly, secure, modular and easy to manage. Include video preview, manual editing of titles/descriptions, scheduling, retry functionality and generation/publishing logs.

I understand that completely unlimited cloud AI generation is not realistically free. Therefore, prioritize local/open-source tools so the automation itself can be unlimited, with optional paid APIs only where necessary.

Before development, please provide the proposed architecture, tools, server/GPU requirements and development plan. Do not add unnecessary features or change the main concept.