The Best Virtual Staging APIs in 2026: 10 Providers Compared

A virtual staging API lets your own product send a room photo and get back a furnished, restyled or relit version, without your users leaving your app. That suits listing portals, rental marketplaces, photography platforms, property-management software, furniture retailers and interior-design apps. This comparison covers ten APIs as their public pages stood in early October 2026.

We make one of them, Roomagen, so read its section with that in mind. Every price below links to the vendor page it came from, and where a vendor does not publish something, the tables say so instead of guessing.

The short answer, by constraint:

• Lowest published price: ModelsLab, at $0.02 per interior render inside a plan from $21 a month. InstantDeco follows at about $0.05 per photo on annual billing.
• Simplest billing: Decor8 AI, $0.20 per image for every API operation at its default output size, with no subscription.
• Staging, day to dusk and floor plan to 3D on one key, prepaid: Roomagen, at $0.20–$0.25 per image in packs that never expire, with failed jobs refunded.
• White-label output plus SSO for brokerages: AI HomeDesign, at $0.245 per photo billed yearly.
• Listing video and 360° tours next to staging: Pedra, priced in euros.

If you are an agent or designer looking for an app rather than an API, our comparison of virtual staging software is the better guide.

10 Virtual Staging APIs at a Glance

How we compared them. We read each provider's pricing page and API documentation on October 1–2, 2026, and ran no blind quality test. Prices are in USD unless marked, exclude tax and change often, so recheck before you sign. "Not stated" means the public pages do not say. The list covers AI APIs with public documentation that can stage or restyle rooms, plus Autoenhance.ai, an enhancement-only API developers often compare with them.

Self-Serve APIs Billed per Image: Roomagen, Decor8 AI and ModelsLab

Roomagen

Roomagen's API has been self-serve since August 30, 2026: you create a key in the developer portal without a sales call, as the quickstart shows. This section sticks to what the developer docs document.

What is documented. One job endpoint, POST /v1/jobs, runs one tool on one image, sent as a public HTTPS URL (JPEG, PNG or WebP, up to 20 MB) or as base64. The tools page documents every tool the API accepts, 65 as of October 2026, including:

• virtual-staging: empty-room, restyle and recreate modes, 13 styles plus a custom option, and 15 room types. It always renders at 2K.
• virtual-renovation, image-enhancement and item-removal.
• day-to-dusk: time of day (0–23), window view, season, and a flag for photos without windows.
• floor-plan-to-3d: documented since September 30, 2026.

Every slug GET /v1/tools returns is documented there, with its options and price.

Pricing (docs):

• 500 images: $125 ($0.25 each)
• 2,500 images: $550 ($0.22 each)
• 10,000 images: $2,000 ($0.20 each)
• 50,000 or more: by arrangement
• 50 free images arrive with your first key, once per account; renders carry a Roomagen watermark until you buy a pack

Packs never expire and they stack. A 4K render counts as two images, and a job that fails during generation is refunded automatically.

Delivery. The docs say jobs are typically ready in 20–60 seconds; staging sits at the slow end, with a median of about 45 seconds in our API logs for the 30 days to October 1, 2026. Results arrive by polling or a webhook signed with HMAC-SHA256 and delivered in up to five attempts over about 36 minutes. Each key can submit 60 jobs a minute, and an Idempotency-Key header makes retries safe. Renders carry a Roomagen watermark until the account buys a pack; after that, output carries no Roomagen branding or watermark. It is served from Roomagen storage under retention limits, so download and keep it. The docs give code samples in cURL, Node.js and Python.

Strengths: One key covers AI virtual staging, virtual renovation, AI photo enhancement, day to dusk photo editing, object and item removal and a floor plan to 3D dollhouse view. Failed jobs cost nothing, and prepaid packs do not expire.

Limitations: The API is young, with no public track record at high sustained volume. There is no batch endpoint: you submit one job per image. The in-app batch mode and the Refine editor are not part of the API. It adds no disclosure label and returns no "original plus edited" pair, so your platform handles disclosure. At $0.25, the entry pack costs more per image than InstantDeco or ModelsLab.

Decor8 AI

Decor8 AI charges $0.20 per image for all API operations, pay as you go. Its Enterprise tier offers custom pricing with volume discounts through sales. Its API reference adds that the rate covers images up to 1,536 pixels on the long side, the default, and that larger outputs from the staging and upscale endpoints use extra credits.

The reference lists room staging, inspirational designs, kitchen and bathroom remodels, wall and cabinet colors, wall priming, landscaping designs, object removal, upscaling, a sketch_to_3d_render endpoint that turns sketches and floor plans into 3D room renders, and replace_sky_behind_house with day, dusk or night skies. A webhooks_data parameter posts results to your URL when generation completes. There are SDKs for Python, JavaScript and Dart (GitHub), plus WordPress and Bubble plugins; the Zapier integration is marked "Coming Soon" on its integrations page.

Strengths: The simplest billing here, a wide list of operations, and official SDKs in three languages.

Limitations: The dusk option replaces the sky behind a house; the docs do not describe relighting windows or interiors. White-label terms and failure policy are not stated, outputs larger than 1,536 pixels cost extra credits, and the base rate only drops through an enterprise negotiation.

ModelsLab

ModelsLab's Interior API prices each interior render at $0.02, drawn from the usage your plan includes: $21 a month on Basic (about 1,050 renders) and $47 on Standard (about 2,350). Its interior models are self-hosted, so the $149 Open Source Unlimited plan has no per-render charge on them.

The page lists room restyling, virtual staging, floor plan generation, a sketch-to-floor-plan endpoint, furniture replacement and 3D walkthrough renders, and says enterprise plans are available for white-label tools.

Strengths: The lowest published price per render in this comparison, and floor plans from sketches in the same API.

Limitations: The interior API is one of many model APIs ModelsLab sells, and its product page does not describe webhooks or failure refunds. At this price, test output quality on your own photos before you plan around it.

Subscription APIs: InstantDeco, AI HomeDesign and Pedra

InstantDeco

InstantDeco's API page offers two plans, and one API call equals one photo:

That is about $0.05 per included photo on Professional billed annually and $0.16 on Starter billed monthly, and the page says "No commitment. Cancel at any time."

Requests are asynchronous: a webhook callback returns the final image. Transformation types include furnish (which leaves blocked elements such as walls, floors and windows unchanged), renovate, redesign, outdoor, blue sky and day_to_dusk, across 10 design styles. The plans list 4K downloads and relighting.

Strengths: The lowest per-photo price here after ModelsLab, a monthly option, and day to dusk included.

Limitations: White-label is in the page title, but its terms are not spelled out, and no failure or retry policy is published.

AI HomeDesign

AI HomeDesign's B2B page sells tiers of 200, 1,000, 2,000, 4,000 and 10,000 photos a month. Billed yearly, every tier works out to $0.245 per photo ($588 a year for 200 photos a month). Billed monthly, it is $0.495 ($99 a month for 200). The page says yearly billing saves 51%.

One REST API with webhooks covers virtual staging, image enhancement, day to dusk, item removal and renovation. The FAQ states that "Outputs return with no AI HomeDesign branding or watermark", and signing up gives free credits with no credit card. The page also lists SSO, credit rollover and "MLS compliance built-in", which is the vendor's claim.

Strengths: The only explicit no-branding statement in this group, webhooks, free test credits and credit rollover.

Limitations: The headline $0.24 rate needs a yearly commitment, $588 at the smallest tier. On monthly billing the price per photo doubles.

Pedra

Pedra sells credits in euros: monthly plans run from 100 credits for €29 (€0.28 each) to 600 for €120 (€0.20), and annual plans from 1,200 credits for €239 (€0.19) to 7,200 for €1,008 (€0.14). A staged or enhanced photo costs 1 credit, and the plans list "API Integration".

Its API specification covers furnishing, emptying and renovating rooms, enhancement with optional perspective correction, blue sky, blur, object removal, prompt-based edits, listing videos and 360° virtual tours. SDKs exist for Python and Node.js, plus an MCP server for AI agents. Image and video calls are synchronous: image endpoints return in about 10–30 seconds, with no job ID and no webhook; only virtual tours are processed in the background and polled. New accounts start with 7 free credits, no card required. A thumbs-down through the feedback endpoint can return the credit for eligible generations (empty room, object removal, prompt edits, renovation and most furnish requests) made within the last two hours.

Strengths: Video and tours next to staging, an MCP server, and a documented credit-back route for bad results.

Limitations: Pricing is in euros, and the published plans stop at 600 credits a month. Synchronous calls hold a connection open for each image, which complicates large queues.

Plan-Gated and Quote-Based APIs: Virtual Staging AI, Collov AI and Gepetto

Virtual Staging AI (Zillow)

Zillow acquired Virtual Staging AI in October 2024, and in September 2025 its technology launched as Virtual Staging inside Zillow Showcase (Zillow). The API page says API "access is included in the enterprise plan", and the pricing page lists that plan at $79 a month with yearly billing ($948 a year) for 150 photos a month, about $0.53 per photo, with a slider for more photos.

The v2 API handles staging, day to dusk with dusk, golden or blue skies, enhancement and furniture removal, and it delivers results by webhook.

Strengths: The staging product Zillow chose for Showcase, current v2 documentation and webhooks.

Limitations: At about $0.53 per photo on its Enterprise tier, it has the highest per-photo rate published here. The pricing page shows monthly billing as more expensive. For portals that compete with Zillow, building a core feature on a Zillow-owned API is also a strategic question.

Collov AI

Collov AI's Enterprise plan costs $127 a month billed yearly ($1,524 a year) for 526 photo credits and includes API access; extra photos cost $0.24 each, and usage resets every month. Monthly billing costs more.

The API docs list staging, room clearing, photo enhancement, chat-based editing, add furniture, material overlay, season change, virtual twilight, lawn replacement and pool effects. Tasks are asynchronous, but you poll for results, and every enterprise API is limited to 10 requests per second and 2 active tasks. The API page promises "No Collov branding".

Strengths: A wide list of visual effects and a clear no-branding statement.

Limitations: Polling only. The 2-task concurrency cap slows bulk jobs, and the low rate needs a yearly commitment.

Gepetto

Gepetto's API page offers a free trial with 50 API calls, and paid pricing is by contact. The documentation lists Furnish, Smart Staging, Creative Redesign, Refresh (walls and floors), Sunshine (sky), Declutter, Upscale, and a Magic Enhancer that can hide people's faces or license plates, plus webhooks. Gepetto says most jobs finish in 15–30 seconds.

Strengths: The 50-call trial lets you benchmark before any sales conversation, and automatic face and plate blurring helps with privacy.

Limitations: With no published price, you cannot forecast cost before talking to sales.

The Adjacent Option: Autoenhance.ai for Enhancement-Only Pipelines

Autoenhance.ai is not a staging API; it is a documented real estate photo API for a different job, enhancement. Its API docs cover HDR bracket grouping and merging, and its enhancement settings cover window pulls, sky replacement, lens and vertical correction, auto privacy, and small "restaging" edits such as fire in fireplaces, grass greening and TV blackout. Webhooks report each processed image. It does not furnish rooms. Its AI transparency page says every AI-processed export carries embedded metadata and C2PA Content Credentials, with an optional visible "AI Modified" badge.

The pricing page offers pay per image, monthly subscriptions and Enterprise, with prices that load in the browser. You can preview enhancements free and pay only for what you download. The plan prices we quoted in August no longer appear, so we removed them.

Fit: Photographer pipelines that fix raw captures at scale. If you also need staging, pair it with a staging API or use one whose operations already include enhancement.

What 500, 2,500 and 10,000 Images a Month Cost

Published prices only, as of early October 2026:

Three things move these numbers more than the sticker price. Billing period: AI HomeDesign's monthly price is double its yearly one, and InstantDeco's monthly plans cost 40–61% more than its annual ones. Resolution: Roomagen counts a 4K render as two images, and Decor8's rate covers outputs up to 1,536 pixels, with extra credits above that. And failures and retries, covered next.

Webhooks, White-Label and the Cost of Failure

Three details rarely make it into pricing tables, but they decide whether an integration survives production.

Delivery. Webhooks let you queue thousands of jobs and react as each one finishes; polling works but costs requests and adds latency. Synchronous calls are the simplest to prototype and the hardest to scale.

Branding. If the images your users download carry a vendor watermark, the API advertises to your customers. AI HomeDesign and Collov state that API output carries no vendor branding; Roomagen's carries none once the account has bought a pack, and renders made before that are watermarked. Elsewhere, read the terms before you launch.

Failures. Generative models sometimes return an unusable image, and at 10,000 images a month that adds up. Separate two cases. A technical failure is refunded automatically by Roomagen and needs a question to most others. A completed but bad result is different: Pedra documents a credit-back route for some operations, Autoenhance charges only for downloads, and at Roomagen a retry costs another image. Most others publish no policy, so ask each provider both questions before you commit volume.

Disclosure Rules Your Integration Must Handle

No API makes a listing compliant on its own, and the rules are getting stricter:

• California sales listings. Since January 1, 2026, agents who advertise with a digitally altered image must comply with Business and Professions Code §10140.8, added by AB 723: a statement that the image was altered, plus a link, URL or QR code to the original. The California DRE's March 2026 advisory adds that responsibility for AI output "rests with the licensee and their responsible broker, not the technology provider."
• California rentals. AB 2025 brings similar disclosure to rental ads from January 1, 2027, covering images changed with editing software or generated entirely by AI, and owners as well as agents (California Apartment Association).
• MLS rules differ on the label. Canopy MLS requires the label on the image itself, not only in the caption (Canopy MLS), while NWMLS does not accept photos with superimposed text or graphics (NWMLS). A label burned in by the API can break one board's rules while satisfying another's.
• European Union. Article 50 of the AI Act has applied since August 2, 2026. Providers of generative systems must mark output in a machine-readable way, and deployers must disclose deep fakes. Under the AI Omnibus, adopted in July 2026, systems already on the market get until December 2, 2026 for the marking duty (guide). Whether an AI-staged photo of a real property counts as a deep fake is not settled (see our EU AI Act Article 50 guide).

What this means for your integration: store the original next to every edited image, keep a record of which operation changed it, and keep label placement under your platform's control so it can follow each market. You already hold the original, because you sent it. Among the vendors here, Autoenhance.ai documents embedded AI metadata and an optional visible label on its exports, AI HomeDesign markets MLS compliance features, and Roomagen's API adds no label. Our AB 723 guide and MLS photo requirements guide cover the board-by-board details.

Build vs Buy: Raw Model Costs vs a Vertical API

Every developer evaluating these prices eventually asks: the models are public, so why pay $0.20 when raw generation costs a few cents?

The raw numbers really are low. Google's Gemini API pricing lists Gemini 3.1 Flash Lite Image at $0.0336 per 1K image, Gemini 3.1 Flash Image at $0.067 (1K), $0.101 (2K) and $0.151 (4K), and Gemini 3 Pro Image at $0.134 (1K or 2K) and $0.24 (4K), with batch requests at about half price. The $0.039 model our August version cited, Gemini 2.5 Flash Image, is deprecated; the same page gives October 2, 2026 as its shutdown date. FLUX.1 Kontext Pro costs $0.04 per image on fal.

Roomagen's image tools run on Google's Gemini image models, so we pay rates like these ourselves. What the spread buys is everything between a model and a product. A raw model does not know that staging an empty living room must keep the windows, doors and flooring, or that redesigning a furnished room should not move the walls. Getting there takes sustained prompt work and evaluation on real room photos. The model also does not retry failures, sign webhooks, dedupe jobs, store results or refund failed jobs; that is code you write, host and maintain.

ModelsLab's $0.02 shows another route: self-hosted models priced below most of the hosted Gemini rates above, a fair trade only if the output holds up on your photos. For teams whose product is a listing, design or retail app rather than image infrastructure, the spread often costs less than the engineering time it replaces. For teams generating hundreds of thousands of images a month with in-house ML staff, building on raw models is a legitimate path, and that is where the math flips.

How to Test a Staging API Before You Commit

Marketing galleries are curated. Quality differences show up on your hardest photos, so test on those:

1. Build a test set from your worst real inputs: dark rooms, wide-angle distortion, cluttered rooms, odd layouts, exteriors at noon.
2. Spend the free allowances first: Roomagen's 50 images, Gepetto's 50 calls, AI HomeDesign's signup credits, Pedra's 7 signup credits and Autoenhance's previews. Decor8 costs $0.20 per test image, and InstantDeco can be tested on a monthly plan.
3. Check structure, not just style: windows, doors, flooring, walls and the view should stay where they were. For listings, a permanent feature that changed is a disclosure and MLS problem, not a cosmetic one.
4. Check consistency across one listing: the same style should read as one home from room to room.
5. Time it end to end, including webhook delivery, and note whether calls are synchronous or asynchronous.
6. Force failures: a dead image URL, an oversized file, an unsupported format. Check the error codes, whether credits come back, and whether a retried request creates a duplicate job.
7. Read the invoice: what did failed jobs, retries and 4K renders actually cost?
8. Confirm the paperwork: output branding, usage rights in the terms, and how long results stay downloadable.
9. Plan disclosure: where originals are stored and who adds which label in each market.

For a code-level walkthrough, see our developer tutorial for adding virtual staging to an app.

Honest Caveats: Where Every Provider Falls Short

A comparison you can trust has to say the uncomfortable parts out loud, including about its author.

• Roomagen: a young API with no batch endpoint, no disclosure features, and a $0.25 entry rate above InstantDeco and ModelsLab.
• ModelsLab: the lowest price, but nothing published on webhooks or failed renders.
• InstantDeco: white-label terms and failure policy are not published.
• Decor8 AI: "dusk" means a new sky, not relit windows.
• AI HomeDesign: monthly billing costs twice the yearly rate.
• Pedra: synchronous image calls with no webhook, euro pricing, plans that stop at 600 credits a month.
• Virtual Staging AI: the highest published per-photo rate, and owned by Zillow.
• Collov AI: polling only and a limit of 2 active tasks.
• Gepetto: a useful trial in front of unpublished pricing.
• Autoenhance.ai: it does not stage, and its prices only show in a browser.

Which Virtual Staging API Should You Pick?

Two notes cut across every row. Room measurements: none of these APIs is built for measured plans. ModelsLab lists a photo-to-floor-plan beta that uses depth estimation, which is an estimate rather than a survey; measured plans come from scan-based capture tools, covered in our floor plan API comparison. Small volumes: if you stage a few dozen photos a month, you may not need an API at all. Roomagen's Starter plan costs $12 a month for 60 credits, or 30 images at 2 credits each, and paid plans can batch 3, 7 or 20 photos per run on Starter, Pro and Business. For more API pricing detail, see our virtual staging API pricing guide.

The Bottom Line

There is no single best virtual staging API in 2026, only a best API for each constraint: ModelsLab and InstantDeco on unit cost, Decor8 AI on billing simplicity, AI HomeDesign on white-label features for brokerages, Pedra on video and tours. Virtual Staging AI and Collov AI bundle their APIs into top-tier plans, Gepetto starts with a trial and a sales call, and Autoenhance.ai solves a different problem.

Roomagen sits in the middle on price: $0.20–$0.25 per image buys staging, renovation, enhancement, day to dusk, item removal and floor plan to 3D behind one key, with failed jobs refunded and packs that never expire. The honest trade-off is track record: its public API only launched on August 30, 2026.

To test it, create a key and follow the API quickstart; your first key includes 50 free images. To see results before writing code, try AI virtual staging or day to dusk photo editing in the browser. The Free plan's 6 one-time credits cover three watermarked runs.