Miko Video Ad Stack

Join 1000+ Marketers Inside Content System AI →

One skill that walks you through making a video ad: brand research → reference breakdown → editable ad plan → Higgsfield generation.

Once the skill is loaded, just say:

Start the video ad stack.

It asks for your brand information, researches in the visible in-app browser, asks for reference videos, shows the breakdown, helps adapt the concept to your product and previews the script and generation plan. Make edits, then say generate when you’re ready.

Setup

Download with Code → Download ZIP and extract, or clone with Git. Open this folder as a local project in an agent with local tools. Connect official Higgsfield separately. Verify visible in-app browser access, ffmpeg/ffprobe, frame inspection and Higgsfield visual/audio analysis. No separate speech-transcription dependency is required. Generation credits are separate. Never paste keys into chat or commit credentials.

Guided walkthrough is the default. You do not need to name internal files or run four separate skills. It requests information when needed, remembers your answers and waits for generation approval. Choose end-to-end explicitly to continue within authorized scope without teaching pauses. Keep campaign outputs outside this public repository.

For native discovery, copy the single skills/video-ad-stack folder, including all its contents, into your local project’s .agents/skills/. Preserve existing skills. Open a fresh task and verify video-ad-stack is available. Then say Start the video ad stack. Cloning alone does not install skills or connect tools. See the official skill guide and local projects guide. Local and cloud projects have different file-access paths.

Alternatively, open the downloaded repository in your local agent and say: “Read the Video Ad Stack skill and start it.” The entry is Video Ad Stack. You do not need to say “follow RUN-STACK.md.”

Preview release: structure validated; real end-to-end campaign pilot still pending. The analyzer incorporates Miko’s Video Breakdown SOP. Both ffmpeg/frame inspection and Higgsfield analysis are required in the stack, plus speech verification when applicable. The planner produces detailed text direction with product-image references at the source target length. This is a skill kit requiring tools and credits, not hosted rendering software.

Before calling this production-tested, complete a real brand/reference pilot and verify target-app setup. Public GitHub files are accessible without joining a community.

The skill contains four internal stages. Brand research and video analysis converge in adaptation. Analysis uses ffmpeg/frame inspection plus Higgsfield visual/audio analysis, without a separate speech-transcription dependency. Generation uses detailed text direction and your product images at the reference’s target duration where supported.

More from Miko

AI Content System — further workflows and resources. Provider credits and private hosted systems are not included.