Michael Aubry

GPT Image 2.6 + Gemini Omni 1.2 is the AI combo of the week.

→ GPT Image 2.6 is the art director: character sheets, first and last frames
→ Omni 1.2 is the camera crew: 15 second scenes, real dialogue, lip-sync
→ One line ("make it anime") restyles the same take

Same faces in every shot, all on one board: wireflow.ai/models/gemini-…

Comment OMNI and I'll send you the board. What would you make first?

I turned Claude into a viral video machine.

3 workflows I actually run:
1. Trend Radar spots what's taking off early
2. Remake Machine remakes proven videos with me in them
3. Content Audit shows why one reel hit 23K

Clone them free: wireflow.ai

Which one first?

My coding agents merged 30 PRs today. My content agents made 11 reels in 5 days. Median: 139 views.

Same agents. Same me. Code has tests, so a bad PR never lands. Content only had my taste, and I stopped checking it.

The one reel that worked last month did 23,441. It named a tool people already know and showed the result in the first second. The 11 that flopped did neither.

So now nothing gets made until we've torn down a post that already won, with its view count next to it. This is the board I check every morning: wireflow.ai/flow/cmw58lodf…

What's your test before a post ships?

Oh boy

Every video tool is shipping an MCP this week. Instagram put one in Edits for Mac, and Higgsfield says it one-shotted a 10 minute doc in Claude.

We've had Claude cutting video through ours for a few months. Biggest lesson: it needs eyes more than it needs tools.

In August it cut a trailer on its own. 58 seconds long, duplicate clips, overlays out of order, and it told us it was fine. It never looked at a single frame.

Now it pulls stills from the cut before it renders and fixes what it sees in plain English. This whole film was made that way. 147 credits, and I never opened an editor.

wireflow.ai/mcp

Anyone letting Claude edit for real yet? Does yours actually look at the frames?