Lampo

Direct your AI video agent. Talk, draw, send feedback. Check the next render. Your taste stays with the project. Building in public.

[Lampo @lampo_video]

@akfullstack “render complete” is a machine status. “ready to publish” is a human decision.

confusing the two is how a typo gets a launch party.

[Lampo @lampo_video]

New chapter here: this account is becoming Lampo.

Your AI video agent doesn't need another paragraph. It needs the right frame.

We're building frame-exact review: point, draw, let the agent fix it, then verify before/after.

Building in public. More soon.

[Lampo @lampo_video]

@akfullstack yep. “looks impressive” and “I’d ship this” are very different verdicts.

we’re building the bit where you can talk or draw over the cut, send that context to the agent, then check what actually changed. generation gets V1. direction gets it out the door.

[Lampo @lampo_video]

@moroboco that makes sense — the source audio is the clock, not a guessed beat sheet.

the tricky revision is “move this one cue” without shifting everything that already lands. useful context for how we’re building Lampo’s feedback handoff.

[Lampo @lampo_video]

@Miguel07Code the focus pull is doing the explaining here. a “make it clearer” revision that just enlarged all the labels would lose the whole idea.

this is why “what should stay” belongs in the feedback too.

[Lampo @lampo_video]

@ProductFaculty the Adobe version reads much clearer. those blue handwritten diagrams do more for me than the dense code panels in the first cut.

6 hours of revisions is the honest part of this thread. “generated” isn’t the finish line.

[Lampo @lampo_video]

@allodev my bet: designers who can direct these agents end up shipping more, not disappearing.

“make a video” and “make the right video” are different jobs.

[Lampo @lampo_video]

the agent: fixed.
me: show me.

lampo keeps the feedback, the next render and the before/after together.

let the agent do the edit. you make the final call.

prepared demo · building in public

[Lampo @lampo_video]

@liu8in A revision benchmark would be interesting next: change one detail, preserve the rest, then have a person approve it. We’re building Lampo so you can talk over the render and send that direction to the agent. How are you testing that second pass?

[Lampo @lampo_video]

@dmytrovirych The rubber-hose brief gives the agent something concrete to protect. Did you need many corrections after V1? We’re building Lampo so project rules and recurring feedback reach the agent before its next render. Keeping the character through V2 is the test.

[Lampo @lampo_video]

@akfullstack Happy to chat here. Are you making videos with Remotion/Claude, or working on client review? We’re building the part between an agent’s first render and a version you’d actually approve.

[Lampo @lampo_video]

We made this teaser with @claudeai (Opus 5.5) + @Remotion, using Seedance footage from @BytedanceTalk.

Reviewed and refined entirely through Lampo: pin the exact frame, send the note, let the agent fix it, check the next render.

[Lampo @lampo_video]

“Around 0:12” is a guessing game.

Meet Lampo. Point at the exact frame. Your AI video agent gets the frame and the note, makes the fix, and sends a new render. You check before/after.

Coming soon.

[Lampo @lampo_video]

@samgrows The first render is getting cheap. The expensive part is “that transition, two frames earlier” turning into five paragraphs. We’re building Lampo around that loop: pin the exact frame, let the agent fix it, verify the next render. Still building in public.

[Lampo @lampo_video]

@SergiiKirianov Wow, congrats! 🤩

[Lampo @lampo_video]

@pradumna_saraf Great choice! 👏

[Lampo @lampo_video]

@SimonJalbert Whenever there is something frustrating you a lot, chances are, other people feel the same way.
And they might be your future customers 😇

That's also why we started #caisy.

[Lampo @lampo_video]

@KarriMarketing Choosing the right CMS is crucial for an enterprise's success. #caisy has special features that make it the best choice for big companies.