Scoring method
Every StageBench score is out of 100 and built from six parts. Every tool gets the same six rooms. We publish the raw outputs, the settings, test dates, and the prices we paid so you can check our work.
The 100 points
- Realism (30 pts). A blind panel of at least three raters, including working real estate agents, rates each output 1–5. Images are shown in random order without tool names.
- Fidelity (25 pts). Did the real property stay the same? We check windows, doors, outlets, flooring, cabinets, walls, and the view out the window. We take off 5 points for each change. If any structural element is added or removed, the result gets a red "Misrepresents property" flag.
- Style quality (15 pts). The blind panel rates fit, furniture scale, and taste from 1 to 5.
- Speed (10 pts). Time from upload to the first result. Under 30 s = 10 · under 2 min = 7 · under 15 min = 4 · slower = 1.
- Cost per photo (10 pts). Based on the cheapest public monthly plan. Under $1 = 10 · under $3 = 7 · under $10 = 4 · more = 1.
- Disclosure features (10 pts). 5 pts for a built-in "Virtually Staged" label or watermark option, and 5 pts for easy export of the original next to the staged image. Why this matters.
Rules we follow
- Paying never changes a score. That covers paid testing, sponsorships, and affiliate commissions alike.
- Same inputs for every tool, with default or recommended settings unless we say otherwise.
- We publish every output, including the bad ones.
- We re-test when a vendor ships a major update, and we date every result.
- Vendors can report errors. We correct mistakes publicly.