tryonlabs · kailashahirwar · Aug 29, 2026 · Aug 29, 2026
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -17,6 +17,13 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Added
 
+#### 👗 Virtual try-on — Google Vertex (`virtual-try-on-001`)
+- **Google Virtual Try-On** (`GoogleVTONAdapter` / CLI `--model google-vton`): first-party Vertex / Gemini Enterprise `recontext_image`
+  - Dedicated person + product try-on; **not** Nano Banana and **not** `GEMINI_API_KEY`
+  - Auth: `GOOGLE_CLOUD_PROJECT` + Application Default Credentials (`gcloud auth application-default login` or `GOOGLE_APPLICATION_CREDENTIALS`)
+  - 1–4 samples; C2PA/SynthID watermark on by default; MCP tool `vton_google_vton`
+  - Docs: `docs/docs/api-reference/google-vton.md`
+
 #### 🧠 Understand + 🎬 Video — NVIDIA NIM (Nemotron Omni, Cosmos 3)
 - **Nemotron 3 Nano Omni** (`NemotronOmniUnderstandAdapter` / CLI `--model nemotron-omni`): hosted NIM chat (`nvidia/nemotron-3-nano-omni-30b-a3b-reasoning`)
   - Image, video, and audio understand; `NVIDIA_API_KEY`; thinking on by default

diff --git a/README.md b/README.md
@@ -16,7 +16,7 @@ API tutorials, configuration, examples, and agent guides live there — not in t
 
 | Category | Highlights |
 |---|---|
-| **Virtual try-on** | FLUX VTO, Nova Canvas, Kling AI, Segmind, Pruna P-Image-Try-On, FASHN, Nano Banana 2 Lite, **Qwen-Image** (API + local), **Muse Image** (composition) |
+| **Virtual try-on** | FLUX VTO, **Google Vertex VTO**, Nova Canvas, Kling AI, Segmind, Pruna P-Image-Try-On, FASHN, Nano Banana 2 Lite, **Qwen-Image** (API + local), **Muse Image** (composition) |
 | **Generate / edit** | Nano Banana family, FLUX.2, GPT Image, Luma Photon, Seedream 5.0 Pro, Ideogram 4.0, Grok Imagine Image, Pruna P-Image / **P-Image-Ideogram** / Edit / Upscale, **Qwen-Image** (API + local), **Muse Image**; local FLUX.2-dev Turbo |
 | **Understand** | Kimi K2.6 / K2.7 Code / K3 (API), Kimi-VL & LLaVA-NeXT (local), **Qwen3.8-Max** (API) + **Qwen3.8-27B** (local) |
 | **Video** | Veo, Sora, Luma Ray 2 + Ray 3.2, Seedance 2.5, Kling 3.0 / Omni / Turbo, Grok Imagine Video 1.5, Gemini Omni Flash, Pruna P-Video / Replace / Avatar / Animate, **LTX-2.5** (API + local), **Hailuo 2.3**, **MiniMax H3** (API + local), **Wan** (API + local 2.2), **Runway Gen-4.5** |

diff --git a/ROADMAP.md b/ROADMAP.md
@@ -31,7 +31,7 @@ This roadmap tracks what shipped and what comes next. Product strategy: [`VISION
 │ • Fashion prompt packs + train-pack schema / datasets                   │
 │ • Train / LoRA recipes (`opentryon train`) + notebooks                  │
 │ • Eval + Fashion Bench v0 (`opentryon eval`)                            │
-│ • One productized local OSS VTON path (CatVTON / IDM / OOT — pick one)  │
+│ • One productized local OSS VTON path (Leffa / CatVTON / IDM / OOT — pick one)  │
 │ • Fashion agentic workflows via MCP (Try-On QA or Fine-Tune Coach)      │
 │ • Efficiency card for that local path (VRAM / latency table)            │
 └─────────────────────────────────────────────────────────────────────────┘
@@ -98,7 +98,7 @@ CLI/MCP remain the distribution spine; new capabilities register the same way.
 - [ ] CLI: `opentryon eval run|report`
 
 ### Slice D — Local OSS VTON (productize one)
-- [ ] Ship **one** of CatVTON / IDM-VTON / OOTDiffusion under `tryon.models` + `opentryon[local]`
+- [ ] Ship **one** of Leffa / CatVTON / IDM-VTON / OOTDiffusion under `tryon.models` + `opentryon[local]` (candidates + licenses: [integrate-next](docs/docs/community/integrate-next.md))
 - [ ] Same invoke path as cloud adapters (agents don’t care where it runs)
 - [ ] Docs: install, VRAM, dry-run, known limits
 
@@ -179,7 +179,7 @@ tryon/
 
 **High-value next:** LoRA recipe hardening, Fashion Bench metrics, first local VTON adapter, Try-On QA or Fine-Tune Coach workflow.
 
-**Vendor / model queue** (not a v0.1.0 commitment): [docs/docs/community/integrate-next.md](docs/docs/community/integrate-next.md) — NVIDIA Nemotron Omni + Cosmos 3 first, then CatVTON, then new services (LipSync, TRELLIS).
+**Vendor / model queue** (not a v0.1.0 commitment): [docs/docs/community/integrate-next.md](docs/docs/community/integrate-next.md) — NVIDIA Wave 1 and Google Vertex VTO shipped; next is OutfitAnyone-Plus and one local VTON weight, then new services (LipSync, TRELLIS).
 
 See [Contributing](CONTRIBUTING.md), [new-model checklist](docs/docs/advanced/new-model-checklist.md), and [VISION.md](VISION.md).
 

diff --git a/SECURITY.md b/SECURITY.md
@@ -58,6 +58,8 @@ KLING_AI_API_KEY
 KLING_AI_SECRET_KEY
 SEGMIND_API_KEY
 GEMINI_API_KEY
+GOOGLE_CLOUD_PROJECT
+GOOGLE_APPLICATION_CREDENTIALS
 BFL_API_KEY
 LUMA_AI_API_KEY
 OPENAI_API_KEY

diff --git a/docs/docs/agents/planner-agent.md b/docs/docs/agents/planner-agent.md
@@ -41,7 +41,7 @@ Each capability has a **default model**. The planner uses that default unless th
 | Video generation | `video` | `video-generate` / `sora` |
 | Background remove | `bg_remove` | `bg-remove` / `ben2` |
 
-- **User named a model** (for example `wan-3.0`, `flux2-pro`, `p-image-ideogram`, `nemotron-omni`, `cosmos3`) → that registry id **only**, even if it lives in another capability. An unknown name is **not** replaced by the default; the planner asks you to pick a real id. `p-image-ideogram` pins Pruna’s model; a bare `ideogram` still pins Ideogram 4.0. `cosmos3-reasoner` pins the VLM, not the Generator.
+- **User named a model** (for example `wan-3.0`, `google-vton`, `flux2-pro`, `p-image-ideogram`, `nemotron-omni`, `cosmos3`) → that registry id **only**, even if it lives in another capability. An unknown name is **not** replaced by the default; the planner asks you to pick a real id. `p-image-ideogram` pins Pruna’s model; a bare `ideogram` still pins Ideogram 4.0. `cosmos3-reasoner` pins the VLM, not the Generator. `google-vton` / `virtual-try-on-001` pin Vertex dedicated try-on, not Nano Banana.
 - **No model named** → the default in the table.
 - The classifier must leave `model` empty unless the user named one. A leaked default in `plan.model` is ignored unless that id also appears in the prompt.
 

diff --git a/docs/docs/agents/vton-agent.md b/docs/docs/agents/vton-agent.md
@@ -12,7 +12,7 @@ keywords:
 
 `VTOnAgent` is a **thin Python facade** over the live registry. It is not a LangChain tool loop. For Studio chat, call [`PlannerAgent`](./planner-agent.md) / MCP `planner_agent`.
 
-Default model: **`kling-ai`**. A named model in the prompt (FASHN, FLUX VTO, Segmind, Qwen-Image, …) pins that VTON registry id.
+Default model: **`kling-ai`**. A named model in the prompt (Google VTO, FASHN, FLUX VTO, Segmind, Qwen-Image, …) pins that VTON registry id.
 
 ```python
 from tryon.agents.vton import VTOnAgent

diff --git a/docs/docs/api-reference/google-vton.md b/docs/docs/api-reference/google-vton.md
@@ -0,0 +1,73 @@
+---
+sidebar_position: 8
+title: Google Virtual Try-On
+description: Vertex AI virtual-try-on-001 — dedicated person + product try-on (not Gemini API / Nano Banana)
+---
+
+# Google Virtual Try-On
+
+First-party Vertex / Gemini Enterprise model **`virtual-try-on-001`**. Person photo + product photo → try-on stills (1–4 samples). Output aspect and resolution match the person image. PNG/JPEG, 10MB max. C2PA / SynthID watermark on by default.
+
+This is **not** Nano Banana composition (`nano-banana-2-lite` / `GEMINI_API_KEY`). The Gemini Developer API does not host this model.
+
+| CLI `--model` | MCP tool | Adapter | Upstream id |
+|---|---|---|---|
+| `google-vton` | `vton_google_vton` | `GoogleVTONAdapter` | `virtual-try-on-001` |
+
+Official docs:
+
+- [Virtual Try-On 001 (Vertex)](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/imagen/virtual-try-on-001)
+- [Generate Virtual Try-On Images](https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/capabilities/generate-virtual-try-on-images)
+
+GA 20 January 2026. Google lists a discontinuation date of 20 January 2027 for this model id — check Vertex docs before relying on it past that.
+
+## Auth
+
+Vertex **Application Default Credentials** plus a GCP project. Studio Connect stores the project id only.
+
+```bash
+gcloud auth application-default login
+export GOOGLE_CLOUD_PROJECT=your-gcp-project
+# optional; default is global
+# export GOOGLE_CLOUD_LOCATION=global
+# or a service account:
+# export GOOGLE_APPLICATION_CREDENTIALS=/path/to/sa.json
+```
+
+Enable billing and the Vertex / Agent Platform APIs on that project. If `global` 404s in your region, set `GOOGLE_CLOUD_LOCATION=us-central1`.
+
+## CLI
+
+No text prompt — the API rejects styling instructions. Shopper photos need `--person-generation allow_adult` (the default).
+
+```bash
+opentryon vton --model google-vton \
+  --person-image person.jpg \
+  --garment-image sweater.jpg
+
+opentryon vton --model google-vton \
+  --person-image person.jpg \
+  --garment-image sweater.jpg \
+  --num-images 2 \
+  --seed 7 \
+  --dry-run
+```
+
+## Python
+
+```python
+from tryon.api import GoogleVTONAdapter
+
+adapter = GoogleVTONAdapter()  # GOOGLE_CLOUD_PROJECT
+images = adapter.generate_and_decode(
+    person="person.jpg",
+    garment="sweater.jpg",
+    number_of_images=1,
+    person_generation="allow_adult",
+)
+images[0].save("worn.png")
+```
+
+## Planner / Studio
+
+Name **`google-vton`** or **`virtual-try-on-001`** in chat to pin this model. The VTON default stays `kling-ai`. After MCP restart, Connect lists **Google Vertex Virtual Try-On** (`GOOGLE_CLOUD_PROJECT`). ADC still lives on the MCP host, not in Studio.
diff --git a/docs/docs/api-reference/nano-banana.md b/docs/docs/api-reference/nano-banana.md
@@ -376,6 +376,8 @@ images[0].save("result.png")
 
 Also available from the `opentryon` CLI / MCP server as `vton --model nano-banana-2-lite`.
 
+This is **composition** try-on via the Gemini Developer API (`GEMINI_API_KEY`). Dedicated Google Cloud Virtual Try-On is `--model google-vton` (Vertex `virtual-try-on-001`, ADC + `GOOGLE_CLOUD_PROJECT`). See [Google Virtual Try-On](google-vton).
+
 ## Command Line Usage
 
 Use the `image_gen.py` script for command-line image generation:

diff --git a/docs/docs/api-reference/nvidia-nim.md b/docs/docs/api-reference/nvidia-nim.md
@@ -14,7 +14,7 @@ One key, `NVIDIA_API_KEY` from [build.nvidia.com](https://build.nvidia.com), unl
 | `cosmos3-reasoner` | `understand` | `Cosmos3ReasonerAdapter` | `nvidia/cosmos3-nano-reasoner` |
 | `cosmos3` | `video-generate` | `Cosmos3VideoAdapter` | Cosmos 3 Generator nano (`POST` infer, `b64_video`) |
 
-Nemotron is **understanding / agents**, not T2I or VTON. Cosmos 3 Generator is **physics-aware T2V / I2V**. There is no NVIDIA VTON NIM — local OSS VTON stays on the [integrate-next](../community/integrate-next.md) Wave 2 list.
+Nemotron is **understanding / agents**, not T2I or VTON. Cosmos 3 Generator is **physics-aware T2V / I2V**. There is no NVIDIA VTON NIM — dedicated try-on APIs and local weights stay on the [integrate-next](../community/integrate-next.md) Wave 2 list.
 
 Official docs:
 

diff --git a/docs/docs/api-reference/overview.md b/docs/docs/api-reference/overview.md
@@ -196,6 +196,33 @@ See [Kling AI API Documentation](kling-ai) for complete details.
 
 ---
 
+### `GoogleVTONAdapter`
+
+Dedicated Vertex AI Virtual Try-On (`virtual-try-on-001`). Person + product images; **not** Gemini API / Nano Banana.
+
+```python
+from tryon.api import GoogleVTONAdapter
+
+adapter = GoogleVTONAdapter()  # GOOGLE_CLOUD_PROJECT + ADC
+
+images = adapter.generate_and_decode(
+    person="person.jpg",
+    garment="sweater.jpg",
+    number_of_images=1,
+)
+```
+
+**Parameters:**
+- `project` (str, optional): GCP project. Defaults to `GOOGLE_CLOUD_PROJECT`
+- `location` (str, optional): Vertex location. Defaults to `GOOGLE_CLOUD_LOCATION` or `global`
+
+**Methods:**
+- `generate_and_decode(person, garment, ...)` - Generate and decode to PIL Images
+
+See [Google Virtual Try-On](google-vton) for auth, CLI, and MCP.
+
+---
+
 ### `AmazonNovaCanvasVTONAdapter`
 
 Adapter for Amazon Nova Canvas Virtual Try-On through AWS Bedrock.