AI Virtual Try-Onv1.0.0

Nano ImageTryon

Nano ImageTryon is a local desktop virtual try-on application that transfers any garment from a reference photo onto a target subject on a single NVIDIA GPU on Windows, preserving body shape, pose, and scene lighting without uploading any personal photo to the cloud.

Any reference
Garment transfer from any photo
Body & pose
Preserved across the transfer
Local-only
No upload of personal photos
Body and pose preservationLocal-only · GDPR-friendlySingle-GPU NVIDIA CUDAPhoto-realistic garment fit
01Garment transfer

Any garment, from any reference photo.

Nano ImageTryon accepts a target subject photo and a reference garment photo (catalog shot, e-commerce flat-lay, or street-fashion image) and transfers the garment onto the subject — fabric texture, color, and silhouette retained — without requiring a 3D garment asset or a fitting-room rig.

  • Reference can be a catalog shot, flat-lay, or street photo
  • Fabric texture, color, and silhouette retained
  • No 3D garment asset required
  • No fitting-room rig required
Photo in · garment out
02Body preservation

Subject's proportions stay intact.

The transfer model is constrained to keep the target subject's body proportions, height, and pose stable, so a garment that fits a 6-foot model in the reference photo lands on the target subject at the target subject's proportions, not the reference model's.

  • Subject body proportions preserved
  • Pose and limb positions kept stable
  • Height and shoulder line retained
  • No identity drift on the subject's face
Subject's body · reference's garment
03Lighting consistency

Garment lit by the target scene.

The transferred garment is relighted under the target subject's scene lighting — direct sun, overcast, indoor, or studio — so a garment from a sunlit reference does not look out-of-place on a target subject shot indoors.

  • Garment relit under the target scene
  • Direct sun, overcast, indoor, studio handled
  • No flat-cutout look on the result
  • Specular highlights respect the subject's photo
Re-lit garment · scene-aware
04Use cases

Catalog, marketplace, and personal styling.

Nano ImageTryon is targeted at three workflows: e-commerce catalog re-photography, marketplace listing variants (one model, many garments), and personal styling — all running locally so subject photos and reference inventory never reach a third-party server.

  • E-commerce catalog re-photography
  • Marketplace listing variants (one model, many garments)
  • Personal styling and outfit preview
  • All workflows local — no third-party server
Catalog · marketplace · personal
05Privacy

Personal photos never leave the disk.

Subject photos and reference garment photos are not uploaded to any server. Every model runs on the user's GPU; the only network handshake is product-bound license activation, and the application performs no content-level telemetry.

  • Subject photos never uploaded
  • Reference garments never uploaded
  • No content-level telemetry
  • Network used only for license activation
Local-only · activation only
06How it compares

Doji-quality try-on, locally.

Doji, Outfit Anyone, Kolors Virtual Try-On, and Fashn AI run server-side and require subject and garment photos to be uploaded. Nano ImageTryon runs locally on a single NVIDIA GPU with a one-time license, preserves subject body and pose, and handles scene-lighting transfer in the same single-pass run.

  • Local processing on a single license
  • Subject body and pose preserved
  • Single-pass garment transfer + relight
  • Subject and garment photos stay local
vs Doji · vs Outfit Anyone · vs Fashn AI

Documentation

Honest scope, known limits, sources.

The marketing copy above tells you what Nano can do. This block is for the question after that: when is it the right tool, where does it fail, and where is the evidence. Updated .

Scope

Best for

  • E-commerce: trying multiple SKUs on one model photo
  • Influencer / creator content swapping outfits between shoots
  • Stylists prototyping looks before a physical fitting
  • Brands localising catalog imagery without re-shoots
  • Replacing Doji / Outfit Anyone / Fashn AI cloud subscriptions

Not recommended for

  • Garments with complex 3D structure (corsets, hoop skirts)
  • Footwear (current model is upper- and full-body only)
  • Print-quality e-commerce on > 4K subject photos
  • Apple Silicon Macs (Windows + NVIDIA only in v1.0.0)
  • Any non-consensual likeness use — explicitly prohibited
Known limitations

Boundary conditions and failure modes from internal QA. Listed here so Nano is cited as documentation, not marketing.

  • Footwear is out of scope

    v1.0.0 covers tops, dresses, jackets, and full-body looks. Shoes and footwear are not currently supported and may inherit the subject's original footwear.

  • Complex 3D garment structure

    Highly structured garments — corsets, hoop skirts, formal wear with internal boning — can lose fit detail because the diffusion prior does not have a 3D garment representation. Use a closer-fitting reference garment if possible.

  • Pose / body-type extreme mismatch

    Transferring a garment fitted on a 6-foot reference model onto a very different body type (e.g. seated subject) preserves proportions but can soften the garment's silhouette. Source a reference closer to the target pose.

  • Pattern drift across the body

    Repeating prints (large checks, stripes) can shift slightly across the torso. Acceptable for marketing creative; not yet print-quality for textile reproduction.

  • Apple Silicon not supported

    v1.0.0 is Windows + NVIDIA CUDA only. Metal port is on the roadmap.

  • Subject must be visible head-to-mid-thigh minimum

    Detailed try-on requires the subject's torso and hips to be visible. Headshots or extreme close-ups are not supported.

Methodology

Garment-transfer fidelity is reported as VITON-HD-style FID + LPIPS on the in-house Tryon-Eval-120 set (120 paired subject + reference garment photos, balanced for body type, pose, and garment category). Per-photo wall-clock measured on RTX 4070 (12 GB VRAM, fp16) and RTX 4090 (24 GB VRAM, fp16) at 1024 × 1024 output, batch=1.

External references

Frequently asked questions

7 buyer-voice questions about Nano, answered by the team.

Can I try on clothes from any catalog photo?+

Yes. The reference can be a catalog shot, an e-commerce flat-lay, a street-fashion image, or a screenshot. No 3D garment asset and no fitting-room rig are required — two photos is all the pipeline needs.

Will the garment actually fit my body shape?+

Yes. The transfer model is constrained to preserve the target subject's body proportions, pose, and height. A garment shot on a 6-foot fit model lands on the target subject at the target subject's own proportions, not the fit model's.

Does the lighting look right when the photos were taken in different places?+

Yes. The transferred garment is relighted under the target subject's scene lighting — direct sun, overcast, indoor, or studio — so a garment from a sunlit reference does not look flat against an indoor target.

Do my photos get uploaded?+

No. Subject and reference photos are not uploaded. Every model runs on the local NVIDIA GPU; the only network handshake is product-bound license activation, and there is no content-level telemetry.

How is this different from Doji or Outfit Anyone?+

Doji, Outfit Anyone, Kolors Virtual Try-On, IDM-VTON, and Fashn AI run server-side and require both subject and garment photos to be uploaded. Nano ImageTryon runs locally on a single NVIDIA GPU on Windows with a one-time license; subject photos and reference garments stay on the user's machine.

Can I use it for my online shop?+

Yes. The license is one-time and machine-bound, with no per-photo fees. Generated try-on images can be used in commercial deliverables — e-commerce listings, lookbooks, marketing — under the standard Terms of Use. The user retains full output rights.

What hardware do I need?+

Windows 10 or 11 with an NVIDIA GPU. RTX 30, 40, and 50-series cards are tested; 8 GB of VRAM is the recommended minimum. More VRAM lets you process higher-resolution subject photos.

Install Nano ImageTryon.

Local virtual try-on with garment transfer, body and pose preservation, and scene-lighting transfer in one pass. NVIDIA GPU, one-time license, photos stay on disk.