Watch on YouTube: https://www.youtube.com/watch?v=QOqEUXaXMsk
Channel: 阿石OMP
Introduction
The video from 阿石OMP presents a fully automated AI‑driven video production pipeline that aims to remove the manual editing step from content creation. By combining a virtual AI agent (ZOComputer) with familiar tools such as OBS Studio, Whisper, Hyper, and GPTimage, the creator records raw footage and lets the system handle cutting, effects, captioning, thumbnail generation, publishing, and even repurposing long videos into short clips. The claimed outcome is a rise in monthly YouTube ad revenue from roughly 2‑3 k HKD to about 5 k HKD while the creator’s active involvement is limited to recording.
What the Video Covers
The walkthrough explains each component of the workflow:
- How the AI agent receives commands via messaging and can control a virtual computer to edit video files.
- Why OBS Studio is configured to output separate webcam and screen streams, giving the AI two distinct video feeds for easier processing.
- The use of three hand‑clap markers as a signal for mistakes, prompting the AI to automatically cut those segments.
- Running Whisper on the recorded audio to produce a Cantonese transcript that feeds caption generation and content understanding.
- Employing Hyper (an open‑source animation tool) to create on‑screen graphics from HTML/CSS instructions generated by the AI.
- Using GPTimage (or a similar service) to produce thumbnail options and run A/B tests via YouTube’s built‑in feature.
- Allowing the AI agent to write titles, descriptions, and publish the final video to YouTube, a blog, and social media platforms.
- Splitting each long video into three 60‑second shorts for Instagram/Facebook to drive traffic back to the main channel.
- Revenue impact: the creator states ad income increased from ~2‑3 k HKD/month to ~5 k HKD/month after adopting the AI workflow.
- Upcoming events: an AI content workshop on 14 Nov 2024 and an OMP event on 19 Nov 2024, both promoted with early‑bird discounts.
Step‑by‑Step Walkthrough
- Set up the AI agent (ZOComputer) and link it to a ChatGPT account (noted as a $100 subscription in the notes).
- Install OBS Studio and configure two virtual cameras: one for the webcam feed and one for the screen capture.
- Record the video. Whenever a mistake is made, clap three times loudly; the AI detects this pattern and marks the segment for removal.
- After recording, run Whisper on the audio track to obtain a text transcript in Cantonese.
- Feed the transcript to the AI agent, which uses it to:
- Generate accurate captions.
- Identify sections to cut based on the hand‑clap markers.
- Produce natural‑language instructions for on‑screen animations.
- Pass the animation instructions to Hyper, which renders HTML/CSS‑based graphics that are overlaid on the video.
- Send the edited video (with captions and animations) to GPTimage to create thumbnail candidates; request A/B test variations for YouTube.
- Let the AI agent compose the video title, description, and tags, then publish the final version to YouTube, the creator’s blog, and linked social accounts.
- Take the original long video and automatically generate three 60‑second shorts tailored for Instagram/Facebook, each ending with a call‑to‑action that directs viewers back to the full YouTube video.
- Monitor performance; the AI can iterate on thumbnail tests and adjust metadata based on early engagement data.
Tools and Numbers Mentioned
- ZOComputer – AI agent with a virtual computer interface.
- ChatGPT – referenced as a $100 cost (presumably for API usage).
- OBS Studio – used to split webcam and screen streams.
- Whisper – speech‑to‑text model for Cantonese transcription.
- Hyper – open‑source tool for HTML/CSS‑driven animations.
- GPTimage – thumbnail generator that can output A/B test variants.
- YouTube’s native A/B testing feature – for evaluating thumbnail performance.
- Ad revenue claim: increase from approximately 2‑3 k HKD per month to about 5 k HKD per month after implementing the workflow.
- Workshop date: 14 November 2024 (hands‑on AI content training).
- OMP event date: 19 November 2024 (additional AI‑marketing insights).
- Early‑bird discount – mentioned for both events.
Caveats and Hype Check
The video’s claim of ad revenue jumping to 5 k HKD/month lacks independent verification and should be treated as an anecdotal result rather than a guaranteed outcome. The presentation heavily promotes the upcoming paid workshop and OMP event, indicating a clear monetization motive. Viewers should assess whether the described toolchain fits their technical comfort level and budget before investing time or money.
Concrete Next Actions
- Research and set up a virtual AI agent capable of receiving text commands (e.g., explore ZOComputer or similar platforms).
- Create a ChatGPT API account and allocate the noted $100 budget for initial testing.
- Download OBS Studio, configure two virtual cameras, and test the separation of webcam and screen feeds.
- Record a short test video, using three distinct hand claps to mark errors, and verify that the AI agent can detect those markers.
- Run Whisper on the test audio to confirm accurate Cantonese transcription.
- Experiment with Hyper to generate simple HTML/CSS overlays based on AI‑produced instructions.
- Use GPTimage (or a comparable thumbnail tool) to produce a set of thumbnail options and enable YouTube’s A/B test feature.
- Allow the AI agent to draft titles, descriptions, and tags, then publish the video to your chosen platforms.
- Produce three 60‑second shorts from the long video, each with a clear call‑to‑action linking back to the full upload.
- If interested in deeper training, register for the AI content workshop on 14 Nov 2024 (bring a laptop) and consider the OMP event on 19 Nov 2024 for additional insights.
Leave a Reply