Architectural images carry hundreds of decisions. Framecheck is a prototype for reviewing one render with Claude and returning a short list of details for the designer to check.
Explore the intended workflow on one of the supplied renders. The example notes below were written to show the report format; they are not Claude output.
“Review the windows, lighting, and desk area for details worth checking before presentation.”
A review would organize possible observations and suggest what to verify. Select a topic to see how one item could read.
Compare the frame spacing and shutter positions across the two openings. A rendered view alone cannot confirm whether a difference is intentional.
Verify: Check both window assemblies in the source model and elevations.
This is a manually written demonstration of the workflow, using a real render. No Claude API call produced these notes, and no defects are asserted. Live analysis is still awaiting configuration.
These architectural images were supplied for the prototype. The captions describe possible review focus areas, not AI findings.
These renders show the visual range we want to test once live analysis is configured.
The interface is ready for a one-image review. Live Claude analysis is available only when the prototype's server access has been configured.
Your observations will appear here. Claude will flag possible issues and suggest what to verify in the source model.
Late visual fixes cost time. Framecheck is being developed as a focused review layer for architectural presentation images: identify a possible issue, explain it, and leave the final judgment with the designer.
In the planned live flow, a designer selects a rendered view or architectural image with an optional focus question.
Once configured, the server forwards the image to Claude for observations about possible inconsistencies in geometry, materials, and placement.
Each observation suggests what to verify. The designer decides whether it is a mistake or an intentional choice.
The landing page, illustrative walkthrough, render gallery, and server integration code are built. Live Claude analysis has not been enabled yet; the walkthrough notes are manually authored.
Next, we plan to compare real model observations with a designer's review, track false positives, and measure the API cost per image. We have not set pricing or published pilot results.
Framecheck is intended for independent architects, visualization specialists, and studios preparing images for presentation. The proposed workflow is one image at a time, with a designer checking each AI observation against the source model.
After the live review flow is tested, a limited pilot could help evaluate whether this is useful in a real project workflow. The commercial model and pricing are still undecided.
Serra Gonca Studio is developing Framecheck as a tool for architects and visualization teams reviewing presentation images. Framecheck is the product name; Serra Gonca Studio is the studio behind it.
This is an early prototype. The public page includes architectural renders and an upload interface; live Claude analysis is not currently enabled.
Choosing an image only creates a local preview in your browser. While live analysis is disabled, this page does not submit your image for review.
If live access is enabled, submitting sends a resized JPEG version of the image, your optional focus question, and your access code to the Framecheck server. The server checks the code and forwards the image and question to Anthropic's Claude API. The access code is not forwarded to Anthropic.
Framecheck's application code does not write uploaded images or reports to permanent storage. Hosting and API providers may process request data under their own policies. Do not submit confidential client material without permission. See Anthropic's privacy information.
Framecheck is an early exploration of how AI assisted review might fit into an architect's process. We are developing the concept one practical check at a time.