We built LandscapioAI because getting a realistic budget for even a small backyard redesign still involves (a) paying a designer $1–3 k or (b) guesstimating from Pinterest photos and hoping Home Depot pricing is close. Both are bad for DIYers and for pros quoting jobs.
What it does
Upload a photo (or use satellite view).
Pick style + rough budget.
In ~60 s you get:
an image-to-image redesign (before/after)
a line-item material + labour budget with upper / lower bounds (we target ±15 % accuracy)
a contractor-ready PDF brief (zones, materials, plant list, sqft & cubic-yard calcs)
How it works
Vision stack – Veo 3 for the render; Segment-Anything for yard boundaries; a small ControlNet for consistent hard-scape outlines.
Cost engine – LLM (Mixtral 8x22B, fine-tuned) that pulls region-specific price data from our internal crawler (250 U.S. zip codes so far) and checks against RSMeans public feeds.
Latency / cost – Avg cold start 71 s on a single A10 GPU; warm 24 s. Each render + estimate costs us $0.07.
Front end – Next.js + tRPC; everything behind Cloudflare → S3-signed URLs for assets.
Data we store – Input images (expire after 14 days), anonymised cost queries, and opt-in e-mails. No geolocation stored past the session.
Why ±15 % matters
Contractors typically pad bids by 20–30 % because customers have no baseline. Even if our output is imperfect, it narrows the expectation gap enough that both sides waste less time on re-quotes.
What’s free vs paid
Free plan: 3 designs + 5 revisions per day, standard-res images.
Paid starts at $9/week (hi-res, unlimited, priority GPU queue).
Open questions / things we’re iterating
Extending the cost dataset outside the U.S. (any sources for EU wholesale pricing would help).
Adding slope/topography handling — right now we assume flat ground.
Local-first processing for privacy-sensitive enterprise pilots (garden-center chains).
IhusanAdam•7mo ago
We built LandscapioAI because getting a realistic budget for even a small backyard redesign still involves (a) paying a designer $1–3 k or (b) guesstimating from Pinterest photos and hoping Home Depot pricing is close. Both are bad for DIYers and for pros quoting jobs.
What it does Upload a photo (or use satellite view).
Pick style + rough budget.
In ~60 s you get:
an image-to-image redesign (before/after)
a line-item material + labour budget with upper / lower bounds (we target ±15 % accuracy)
a contractor-ready PDF brief (zones, materials, plant list, sqft & cubic-yard calcs)
How it works Vision stack – Veo 3 for the render; Segment-Anything for yard boundaries; a small ControlNet for consistent hard-scape outlines.
Cost engine – LLM (Mixtral 8x22B, fine-tuned) that pulls region-specific price data from our internal crawler (250 U.S. zip codes so far) and checks against RSMeans public feeds.
Latency / cost – Avg cold start 71 s on a single A10 GPU; warm 24 s. Each render + estimate costs us $0.07.
Front end – Next.js + tRPC; everything behind Cloudflare → S3-signed URLs for assets.
Data we store – Input images (expire after 14 days), anonymised cost queries, and opt-in e-mails. No geolocation stored past the session.
Why ±15 % matters Contractors typically pad bids by 20–30 % because customers have no baseline. Even if our output is imperfect, it narrows the expectation gap enough that both sides waste less time on re-quotes.
What’s free vs paid Free plan: 3 designs + 5 revisions per day, standard-res images.
Paid starts at $9/week (hi-res, unlimited, priority GPU queue).
Open questions / things we’re iterating Extending the cost dataset outside the U.S. (any sources for EU wholesale pricing would help).
Adding slope/topography handling — right now we assume flat ground.
Local-first processing for privacy-sensitive enterprise pilots (garden-center chains).
Would love feedback on:
data-quality / accuracy scepticism (we expect it!)
architectural bottlenecks you see
whether the UX gets out of your way fast enough
Thanks for taking a look — will be here all day to answer questions.
lolbert291•7mo ago