Introducing companion mode The difference between ordering a video and directing one Now your agent can pitch five angles, check in with a storyboard, then sketch frames to review You steer every step and only build when approved $ npx hyperframes@latest skills

The repo:

This approach respects both the user’s time and their API budget.

Tried the pitch-first flow today and it saved me so many trial-and-error runs.

Finally, creative direction control inside an AI tool!

The card-based storyboard layout makes reviewing complex scene logic trivial.

Wasting 30 minutes on a bad prompt output is officially a thing of the past.

The whole point was to get approval before the expensive step, not after.

From random coin flips to predictable excellence—what an upgrade!

Creative control without creative friction. That's the balance this tries to hit.

The shift from ordering a video to directing it is exactly what AI tools were missing.

A creative partner that checks in before acting is more useful than one that surprises you.

Integrating this into our agency workflow immediately. Massive time-saver.

This is how AI creative tools should have worked from day one.

It's been a journey to build up to this point. Gotta start somewhere!

CLI-based video direction feels like live-coding your visual story.

Being a creative director vs being an input field. Huge difference in what you ship.

We’ve all felt the pain of waiting 30 minutes for a ruined logo. Never again!

The future of creative AI isn't full automation—it's guided direction.

Game-changer for developer-marketers who need precise video output fast.

No more explaining to clients why the AI changed the character's face halfway through.

Inserting approval gates into automated pipelines is how enterprise teams will actually adopt AI tools.

This solves the classic problem of AI models ignoring key prompt details.

Does the sketch phase support custom aspect ratios like 9:16 and 16:9 simultaneously?

Easily one of the best out now, DM me to help boost 🚀🚀

What's the default resolution output when rendering the final composition pass?

just in time for next week's tech week

The orchestration around generation is harder to build than the generation itself. Worth it.

Directing instead of just ordering a video is a nice shift in control 🎬

Direction beats generation. You steer toward something specific instead of hoping for it.

Steering toward something specific beats hoping the output lands, that's the real upgrade.

Frame-by-frame approval is a game-changer.

It's been a gatekept secret by some of the best people in AI video, but now it's an opensource skill

Co-creation over automated generation any day.

Less coin-flip, more creative direction. That was the whole design goal.

Sounds like a solid plan! I'm curious to see how it all comes together.

More to come. Keep an eye on the skills catalog.

Ididn’t understand this update well

Real-time feedback loops on specific frames make all the difference.

Predictability is key for client work, and this pipeline guarantees it.

The cost of iteration drops dramatically when you aren't compiling heavy visual assets until the final step.

Combining developer tools with video workflows is surprisingly effective!

Code-native creative tooling follows the developer workflow naturally.

No more burning tokens on bad layouts. This is exactly what was missing!

Save your tokens for more

How fast is the local sketch generation compared to the full MP4 compilation step?

This sets a new benchmark for human-AI creative collaboration.

A step-by-step review process guarantees professional visual alignment

How are the intermediate sketch frames rendered locally before final export?