We built SEELE TV to make multi-shot AI video less chaotic
Hi Indie Hackers,
We recently launched SEELE TV, a planning tool and workflow guide for AI video production.
The problem we care about begins when a video has more than one shot. The original brief starts to drift, and revisions become difficult to trace. A clip might look good on its own but still be wrong for the sequence.
SEELE TV lets you describe what the finished video should communicate and turn that into a production handoff for the SEELE Film & CG workspace. The planner runs in your browser. It does not upload media or call a generation model.
Behind the site is a statically exported Next.js application. We keep the content in a typed registry rather than maintaining separate lists for pages, metadata, links, and source information. Each page also states where workflow advice ends and current product capability begins.
We block production releases when the source information or route validation is incomplete. That may sound strict for a content site, but model availability and product details change often. We do not want an old page quietly turning into a product promise.
We are now trying to answer one question: would you use a planning step before opening an AI video generator, or would it feel like extra work?
The interesting conversion challenge is that you’re asking users to adopt a planning step to solve a problem they may only recognize after generation fails. “Brief drift” makes perfect sense to someone who has already experienced multi-shot chaos, but a first-time AI video creator may interpret planning as unnecessary overhead because the generator feels faster. That makes the positioning less about whether planning is useful and more about making the cost of skipping planning immediately tangible. I’d be curious whether showing the failure mode—rather than explaining the workflow—would make the value of SEELE TV click faster.
Agreed — showing the failure mode may explain the value better than describing it. Our view is that planning shouldn’t be mandatory: if someone just wants a finished video, AI can direct the camera through the 3D scene; if they need an exact shot, they can step in for pixel-level control. Which mode would you use first?
The planning vs. friction debate shows up in every creative workflow that scales past one shot. Brief drift is the AI video equivalent of a director who skips the storyboard and wonders why the edits don’t cut together. The generator is the camera — it can only execute what’s already decided. Your planning layer is the pre-production meeting. To your question: yes, especially once someone has regenerated the same clip four times and still can’t explain why it doesn’t sequence right. The time cost of planning is front-loaded; the time cost of skipping it compounds.
Exactly. We think the workflow should split by intent: AI-directed video when speed matters, and precise shot control when the creator has a specific vision. The Director Desk isn’t fully manual — AI operates the camera through our 3D model, while the creator steps in only when needed.
The layered control model is the right call. Most tools force you to choose - either full AI control or full manual. The handoff point is where the real creative decisions live. Is the intent detection automatic or does the creator explicitly declare which mode they're in?
The agent routes to direct generation or fine‑grained control per user intent, while outer‑layer UI lets users make explicit selections without relying on the agent every time.
Both simultaneously is the harder design problem. Not because the logic is complex but because the UI has to tell the user which mode they're in without making them think about it. The moment someone wonders "is the AI deciding this or am I?" you've broken the flow. How do you surface the active control mode without adding friction?
The tension I notice is that SEELE TV is trying to reduce chaos by adding a planning layer before generation — but that layer itself has to earn a place in a workflow people already consider fast.
That makes the “extra work” question more fundamental than it initially sounds.
That’s the line we’re trying to draw. The Director Desk shouldn’t become another manual tool: AI drives the camera in 3D by default, and the creator only intervenes when exact control matters. Would that feel like less friction, or still one step too many?
That depends a lot on whether creators experience the planning layer as setup or as part of creating. Have you seen a clear preference yet?
There isn’t a super clear‑cut preference across all creators yet. One trend we’re observing is that CG folks prioritize shot controllability. Most of them already make videos in Blender or UE5, so control matters a great deal to them. It’s also interesting to see more professional users adopting AI video as a real productivity tool, rather than just a toy. I’m curious if others here have seen the same pattern.
That professional-user signal is interesting. Would be good to hear how it develops — would you be open to sharing the best email to reach you on?
Thanks for your interest! You can reach me at muzuochen@seeles.ai — happy to keep you posted as this develops.
Thanks! I’ve just sent it over.
Looking forward to hearing your thoughts whenever you have a chance.