Virtual Try-On & Developer APIs

Best Virtual Try-On APIs for Developers in 2026

Six virtual try-on APIs for e-commerce developers, ranked on pricing, model quality, multi-model routing, and commercial license terms — updated after Black Forest Labs launched FLUX VTO in May 2026.

By Arthur R. Published June 8, 2026 Updated June 8, 2026 12 min read

Quick answer

There is no single best virtual try-on API for developers; the right choice depends on whether you need a dedicated purpose-built model, a unified routing layer across multiple providers, or the lowest per-image cost. Fashn.ai ranks first for its publicly listed pricing ($0.075 per image on pay-as-you-go, verified June 2026) and purpose-built e-commerce feature set; TryOn API ranks second for developers who want one OpenAI-compatible endpoint routing 14 models across 7 providers without managing separate integrations.

Key takeaways

No single winner: match the API to your product type, latency requirements, and whether you need multi-model routing or single-provider simplicity.
Fashn.ai ranks
TryOn API ranks #2: one OpenAI-compatible endpoint routes 14 models across 7 providers — Kling Kolors, Gemini, GPT-Image-1, FLUX Kontext, and more — under a single credit-based billing system.
Black Forest Labs launched FLUX VTO on May 28, 2026 — it handles up to 4 garments simultaneously and outputs up to 2 megapixels in under 4 seconds.
Replicate's IDM-VTON is the cheapest option (~$0.024/run) but carries a CC BY-NC-SA 4.0 license that prohibits commercial use — rule it out for production storefronts.
Google Cloud's Vertex AI Virtual Try-On API (GA January 2026) was deployed by OTB Group across Diesel and Jil Sander in May 2026; its current endpoint is scheduled for deprecation January 20, 2027.

Virtual try-on models routable via TryOn API across 7 providers (TryOn API landing page, June 8, 2026)

Maximum garments FLUX VTO applies simultaneously per generation (Black Forest Labs blog, May 28, 2026)

There is no single best virtual try-on API for developers; the right choice depends on whether you need a dedicated, purpose-built model, a unified routing layer across multiple providers, or the most favorable per-image economics — and the competitive landscape shifted in May 2026 when Black Forest Labs launched FLUX VTO, adding a fourth credible foundation model to a category that had fewer than three real API-accessible options a year earlier.

Virtual try-on APIs give e-commerce developers programmatic access to AI that synthesizes a photo of a garment on a person, without physical inventory or manual photography. The practical applications range from product page imagery and look-builder tools to mobile dressing rooms and AI shopping assistants. The infrastructure has matured rapidly: Google Cloud’s Vertex AI Virtual Try-On API reached general availability on January 20, 2026, and was deployed by OTB Group — the parent of Diesel, Jil Sander, Marni, and Maison Margiela — across luxury retail in the US and Europe in May 2026. The category now has enough distinct API-accessible models that the central developer decision is no longer “is there anything usable?” but “do I integrate each model separately, or route them all through a single endpoint?”

Fashn.ai ranks first for direct integrations: its per-image pricing is the most transparent of any provider here, its dedicated model lineup is purpose-built for garment visualization, and it has the most complete public documentation. TryOn API ranks second for developers who want a single OpenAI-compatible endpoint routing 14 models across 7 providers — the same routing argument that made OpenRouter compelling in the language-model space, applied to virtual try-on. FLUX VTO (Black Forest Labs, launched May 28, 2026) ranks third on raw model capability, with the best simultaneous multi-garment support currently available. Below we rank all six API-accessible options on verified pricing, license terms, and production readiness.

Why model fragmentation favors a routing layer

The virtual try-on API market gained at least four meaningfully distinct foundation models in a 12-month window: Google Cloud’s VTO-001 reached GA in January 2026, Fashn.ai’s tryon series continued active development, and Black Forest Labs entered with FLUX VTO on May 28, 2026 — all on top of the already-established Kling Kolors. Each model has a different price point, output resolution, garment-type strength, latency profile, and content policy. FLUX VTO leads on multi-garment simultaneous generation; Fashn.ai’s Max endpoint targets brand-quality fidelity on logos and hardware details; Kling Kolors runs at a consistent ~$0.07/image across a well-tested deployment base; Google Cloud prioritizes enterprise compliance posture.

That fragmentation is the structural case for a routing layer — TryOn API’s core value proposition. If your pipeline needs the cheapest model for bulk catalog shots, the highest-fidelity model for hero images, and a fallback provider for uptime resilience, a single-provider direct integration forces you to pick one and absorb its trade-offs everywhere. A router lets you dispatch each request to the model that best fits the task, without re-engineering the integration every time the market adds a new option. The pace of addition — a new foundation-model entrant in May 2026 alone — suggests that flexibility will matter more, not less, over the next year.

Pricing transparency as a practical selection criterion

One useful signal in API selection is how much pricing a provider publishes before requiring a signup or a sales call. On this dimension the virtual try-on market is currently inconsistent. Fashn.ai publishes a complete credit table on its help center: $0.075 per standard try-on generation on pay-as-you-go, $0.30 per premium Max generation, with volume tiers down to $0.0488/image at $1,249/month — all confirmed June 8, 2026. Kling Kolors’ ~$0.07/image pricing is available on fal.ai’s model page and confirmed by third-party resellers. At the other end: Google Cloud’s VTO pricing page redirected to a 404 during research; FLUX VTO’s launch blog states only that it “runs cheaper than comparable systems” with no specific figure; and TryOn API requires signup to view its model rates.

Opaque pricing is not automatically a dealbreaker, but it blocks pre-build cost modeling — real friction when a project needs a business case before procurement can approve it. For teams that require verifiable unit economics before starting a build, Fashn.ai and Kling Kolors via fal.ai are the practical choices today.

The license filter

Every commercial deployment passes through a license check before quality or price becomes relevant. Five of the six APIs in this report carry standard commercial terms. The exception is Replicate’s IDM-VTON, licensed under CC BY-NC-SA 4.0 — a Creative Commons license that explicitly prohibits commercial use. The lowest raw API cost in this comparison ($0.024/run) is therefore irrelevant for production storefronts, paid products, or any revenue-generating deployment. IDM-VTON belongs in research and non-profit contexts only. If you encounter another article recommending IDM-VTON for e-commerce without a commercial-use caveat, treat the omission as a signal about that author’s rigor.

Where the developer API market is heading

The virtual try-on API landscape is moving from “one or two usable options” to a genuinely fragmented multi-model market. FLUX VTO’s May 28, 2026 launch is a structural marker: Black Forest Labs built its reputation on general-purpose image generation before making a dedicated VTO investment, which signals that the category’s addressable market is large enough to warrant purpose-built foundation-model work from players who previously stayed general. OTB Group’s deployment of Google Cloud VTO across Diesel and Jil Sander in May 2026 is an early data point on enterprise adoption at the high end of fashion retail.

For developers, the practical implication is that the API you choose today may not be the best option 12 months from now, and the integration cost of switching is non-trivial. That shifts the risk calculus toward either the provider with the strongest ongoing model investment — unclear today — or the routing layer that lets you swap providers without rewiring your code. The FLUX VTO launch, a new model entering the market with minimal advance notice, is the clearest recent illustration of why the routing argument has more force in this category than in most.

Fashn.ai

Best dedicated VTO API with transparent pricing

Fashn.ai is the most transparent and specialized virtual try-on API for e-commerce developers, with publicly listed per-image pricing and a model lineup built specifically for garment visualization. The standard model, tryon-v1.6, costs 1 credit per generation — $0.075 on pay-as-you-go, verified from the official help center on June 8, 2026. The premium tryon-max endpoint runs at 4 credits per generation ($0.30 on pay-as-you-go), targets professional product photography quality with higher fidelity on logos, fabric, and hardware, and processes in about 50 seconds. Volume tiers reduce the effective rate: $19/month buys 282 credits ($0.0675/image), $249/month buys 4,150 credits ($0.0600/image), and $1,249/month buys 25,594 credits ($0.0488/image). Annual billing saves roughly two months' equivalent spend. The API accepts flat-lay, ghost-mannequin, and on-model garment images; inputs and outputs are auto-deleted after 72 hours; TypeScript and Python SDKs are available. A background removal endpoint is bundled at 1 credit per call. The minimum spend is $7.50 (100 on-demand credits), and there is no free tier.

Pros

Transparent public pricing verified directly from the official help center on June 8, 2026
Dedicated e-commerce model lineup with standard and premium tiers for different quality-cost trade-offs
Auto-deletion of inputs and outputs after 72 hours; TypeScript and Python SDKs available

Cons

No free tier; minimum spend is $7.50
Single provider — no model fallback if Fashn.ai has an outage or a quality regression
Try-on Max at $0.30 per image may be expensive for high-volume catalog workflows

TryOn API

Best for multi-model routing via a single endpoint

TryOn API is positioned as "the OpenRouter for virtual try-on" — a unified API gateway that routes 14 virtual try-on and generative-AI models across 7 providers through a single OpenAI-compatible endpoint, with pay-per-request credit billing. Rather than maintaining separate SDK integrations and billing accounts for each provider, developers send one request and select which underlying model to invoke: dedicated try-on models (including Kling Kolors), image generation models with VTO capability (Gemini, GPT-Image-1, FLUX Kontext), 3D model generation, and AR try-on are all reachable from the same interface. The endpoint fully supports the chat completions format, streaming SSE, and a modalities parameter for image output; a simpler multipart /tryon route is also available. For teams that need to benchmark models, swap between providers without code changes, or hedge against a single vendor's quality regression, a multi-model routing layer for virtual try-on removes the integration tax that would otherwise accompany each new model entry. The FLUX VTO launch in May 2026 illustrates exactly how fast that market moves. Per-model credit pricing is visible after signup; public rates were not accessible on the landing page as of June 8, 2026.

Pros

Routes 14 models across 7 providers through one OpenAI-compatible endpoint
Model-switching without code changes — benchmark or swap as the market evolves
Credit-based pay-per-request billing with no per-seat subscription

Cons

Per-credit pricing is not publicly listed; requires signup to see the model catalog and rates
Routing layer adds an extra hop between your application and the underlying model
Smaller public documentation footprint than direct-provider SDKs

FLUX VTO

Best for multi-garment simultaneous try-on at catalog scale

Black Forest Labs launched FLUX VTO on May 28, 2026, making it the newest dedicated virtual try-on model in this report and the first major foundation-model entry specifically built for catalog-scale garment generation. Its standout capability is handling up to four garments simultaneously on a single model image, supplied as a 2x2 composite grid — a material advantage for multi-item outfit assembly pipelines. FLUX VTO preserves face and pose while rendering logos, prints, stitching, and hardware with high fidelity, and supports garment layering (a shirt under a jacket with realistic fabric interaction). Output resolution reaches up to 2 megapixels in under four seconds via the cloud API; self-hosting is available for sub-second latency. Flat-lay packshot inputs tend to produce cleaner results than on-model reference images. Swimwear and lingerie are excluded under the default moderation policy. The model is available via BFL's own API and through Runware. Per-image pricing is not published on BFL's official blog or Runware's documentation; BFL states it runs cheaper than comparable systems, but no specific rate has been publicly disclosed as of June 8, 2026.

Pros

Handles up to 4 garments simultaneously on one model image — unique in this comparison
High fidelity on logos, prints, stitching, and garment layering
Self-hosting available for sub-second inference at catalog scale

Cons

No publicly disclosed per-image price as of June 2026
Swimwear and lingerie excluded under default moderation policy
Launched May 28, 2026 — limited real-world production track record

Kling Kolors VTO

Best established model with known per-image pricing

Kling v1.5 Kolors Virtual Try-On, accessible to developers through fal.ai's official integration, is one of the most widely deployed virtual try-on models in production. At approximately $0.07 per generated image — confirmed via fal.ai and PiAPI as of June 2026 — it is priced similarly to Fashn.ai's standard tier. Output resolution is 768x1024 pixels. The API accepts upper-body, lower-body, and full-body/dress garment inputs as separate parameters; these are mutually exclusive, so a single request handles one garment category at a time. Failed API tasks do not consume credits. The model has a longer real-world production track record than FLUX VTO and a broad developer ecosystem. Its limitations are lower output resolution than FLUX VTO, and it struggles with garment-type switching (swapping a skirt for pants, for example) and complex layering. The official Kling developer portal (developer.klingai.com) was inaccessible during our research; fal.ai is the most practical access route for most developers.

Pros

Approximately $0.07 per image on pay-as-you-go (confirmed via fal.ai and PiAPI, June 2026)
Well-tested in production with a broad developer ecosystem
Failed API tasks do not consume credits

Cons

Output at 768x1024 px — lower resolution than FLUX VTO's 2 MP ceiling
Struggles with garment-type switching and complex layering
Official developer portal was inaccessible during research; fal.ai is the practical access point

Google Cloud Vertex AI Virtual Try-On

Best for enterprise deployments within the Google Cloud ecosystem

Google Cloud's Vertex AI Virtual Try-On API (model endpoint virtual-try-on-001) reached general availability on January 20, 2026, and has since been deployed in production by OTB Group — the parent of Diesel, Jil Sander, Marni, and Maison Margiela — as part of a luxury fashion clienteling tool launched across the US and Europe in May 2026. The API accepts PNG and JPEG inputs up to 10 MB, generates up to four output images per request matching the input resolution, and includes SynthID digital watermarking and C2PA Content Credentials by default. The rate limit is 50 regional online prediction requests per minute. Pricing is not listed on the model reference pages; the billing documentation redirected to a 404 at time of research, so verify current rates through the Google Cloud pricing calculator or your account team before committing. One operational note worth monitoring: the current endpoint is scheduled for deprecation on January 20, 2027 — an unusually short GA lifecycle. Google Cloud's $300 free trial credit applies broadly to Vertex AI usage and is the practical entry point for testing.

Pros

GA since January 2026; deployed in production by OTB Group luxury brands (Diesel, Jil Sander)
Up to 4 output images per request with SynthID watermarking and C2PA Content Credentials
Google Cloud's $300 free trial credit provides a cost-free entry point for testing

Cons

Public pricing not accessible on official documentation pages — verify via pricing calculator or account team
Model endpoint (virtual-try-on-001) scheduled for deprecation January 20, 2027
Rate-limited at 50 requests per minute per region

Replicate IDM-VTON

Cheapest option — non-commercial license blocks production use

Replicate's IDM-VTON is the cheapest virtual try-on model accessible via API, at approximately $0.024 per run (roughly 41 runs per dollar), confirmed from the Replicate model page on June 8, 2026. It runs on an NVIDIA A100 80 GB and takes about 17 seconds per prediction. The critical constraint for commercial developers is its license: IDM-VTON carries a CC BY-NC-SA 4.0 license that explicitly prohibits commercial use. Running it in a production storefront, charging customers for the output, or embedding it in a paid product violates those license terms. It is a legitimate choice for academic research, non-profit projects, or internal experimentation, but not a viable option for any commercial e-commerce deployment. Replicate was acquired by Cloudflare (deal closed December 2025); as of June 2026, no new enterprise pricing tiers have been introduced.

Pros

~$0.024 per run — cheapest API option in this comparison by a factor of three
No subscription required; pay per use on Replicate's platform
Suitable for academic research, non-profit projects, and internal experimentation

Cons

CC BY-NC-SA 4.0 license: prohibits commercial use — not suitable for production storefronts or paid products
~17 seconds per inference — slowest on this list
Limited to a single SDXL-based model with no dedicated e-commerce feature set

Best Virtual Try-On APIs for Developers in 2026 — comparison
Tool	Pay-as-you-go price	Free tier	License	Best for
Fashn.ai	$0.075/image (v1.6, verified Jun 2026)	No ($7.50 minimum)	Commercial	Dedicated e-commerce VTO API
TryOn API	Credits — rates after signup	Not confirmed	Commercial	Multi-model routing via one endpoint
FLUX VTO	Not published (BFL, Jun 2026)	No	Commercial	4-garment simultaneous; catalog scale
Kling Kolors VTO	~$0.07/image via fal.ai	No	Commercial	Established model; wide ecosystem
Google Vertex AI VTO	Not published — requires inquiry	via $300 trial credit	Commercial	Enterprise / Google Cloud customers
Replicate IDM-VTON	~$0.024/run (verified Jun 2026)	No	Non-commercial only	Research and experimentation only

No virtual try-on API dominates every dimension in 2026. Fashn.ai leads on pricing transparency and e-commerce specialization; TryOn API is the pick for multi-model routing through one endpoint; FLUX VTO (launched May 28, 2026) leads on simultaneous multi-garment generation; Kling Kolors is the most established model at a consistent ~$0.07/image; Google Cloud Vertex AI has the enterprise compliance footprint; and Replicate IDM-VTON is cheapest but commercially unusable under CC BY-NC-SA 4.0. Verify live pricing pages before committing to any provider — Google Cloud's VTO pricing documentation was inaccessible at the time of this report.

Frequently asked questions

Which virtual try-on API has the most transparent pricing for commercial developers?

Fashn.ai publishes a detailed credit table on its help center. The standard tryon-v1.6 model costs $0.075 per generation on pay-as-you-go; the premium tryon-max endpoint costs $0.30 per generation. Both figures were verified from the official help center on June 8, 2026. All other APIs in this comparison either require signup to see rates (TryOn API) or do not publish pricing on their public documentation pages (FLUX VTO, Google Cloud Vertex AI).

Can FLUX VTO apply multiple garments in a single generation?

Yes. FLUX VTO, launched by Black Forest Labs on May 28, 2026, handles up to four garments simultaneously on a single model image, provided as a 2x2 composite grid on a white background. It also supports garment layering — for example, a shirt realistically tucked under a jacket — which most other single-request models do not handle in one call.

What does TryOn API route, and how does it integrate with existing code?

TryOn API routes 14 models across 7 providers — including dedicated VTO models (Kling Kolors), image generation models with VTO capability (Gemini, GPT-Image-1, FLUX Kontext), 3D generation, and AR try-on — through a single OpenAI-compatible endpoint. It supports the standard chat completions format with streaming SSE and a modalities parameter for image output, as well as a simpler multipart /tryon route. Per-model credit rates are visible after signup; the public landing page does not list prices.

Is Replicate IDM-VTON safe to use in a commercial e-commerce product?

No. IDM-VTON on Replicate is licensed under CC BY-NC-SA 4.0, which explicitly prohibits commercial use. It is suitable for academic research or internal non-commercial experimentation, but using it in a production storefront or any paid product violates the license terms.

Is Google Cloud's virtual try-on API going to be discontinued?

The current model endpoint (virtual-try-on-001) is scheduled for deprecation on January 20, 2027 — roughly one year after reaching general availability on January 20, 2026. Google has not announced a successor model publicly as of June 2026. Teams building on this API should monitor Google Cloud release notes for an updated endpoint before that date.

Which virtual try-on API has a free tier?

None of the six APIs in this report offer a true free tier. Google Cloud's Vertex AI Virtual Try-On is the only one testable at no direct cost — through the platform's $300 free trial credit, which applies broadly to Vertex AI usage. All other options require a minimum spend from the first request.

How does TryOn API compare to calling Kling or Fashn.ai directly?

Calling a provider directly — Kling via fal.ai or Fashn.ai — gives you the most direct call path with no routing margin in the pricing. TryOn API adds a layer that lets you switch between 14 models without changing your integration code. The routing approach pays off when you need to benchmark models, want vendor failover resilience, or expect to adopt new models as they enter the market — such as FLUX VTO in May 2026.

How we evaluated

We assessed each API on six dimensions: publicly verifiable pricing taken directly from official documentation, model pages, or API marketplaces on June 8, 2026 (not from marketing pages); commercial license terms; model quality as corroborated across developer documentation and independent deployment reports; multi-model vs. single-model architecture; production track record; and free-tier availability. Fashn.ai pricing was taken from the official help center; Kling Kolors pricing was confirmed via fal.ai's model page and PiAPI. FLUX VTO pricing was not publicly disclosed on BFL's blog or Runware's documentation as of the research date. Google Cloud Vertex AI pricing could not be verified from official documentation and should be confirmed through the Google Cloud pricing calculator before committing. TryOn API per-request pricing is listed only after signup and was not accessible from the public landing page. License terms were taken from the official license declarations on each provider's model or platform page.

Publicly verifiable pricing from official documentation or API marketplaces on June 8, 2026
Commercial license terms — CC BY-NC-SA 4.0 (non-commercial only) disqualifies from production use
Model quality and garment-type coverage corroborated across developer documentation and deployment reports
Multi-model routing flexibility and model-switching without code changes
Free-tier availability and minimum spend requirements

Innvesti may have commercial relationships with some companies mentioned. Our analysis is editorially independent. Figures cited are sourced; where reliable data is unavailable we say so rather than estimate.

Key takeaways

Why model fragmentation favors a routing layer

Pricing transparency as a practical selection criterion

The license filter

Where the developer API market is heading

Fashn.ai

Pros

Cons

TryOn API

Pros

Cons

FLUX VTO

Pros

Cons

Kling Kolors VTO

Pros

Cons

Google Cloud Vertex AI Virtual Try-On

Pros

Cons

Replicate IDM-VTON

Pros

Cons

Frequently asked questions

How we evaluated

Sources