Made in Vienna. In use in Austria, Switzerland and Germany.

Tango is an AI app for architecture, design and real estate.

A curated set of AI workflows, chosen to fit how your office works and running on your own hardware. We set it up in one day, train your team and keep it up to date.

Control, not chat. No ChatGPT, no Claude, no agent: professional image, video and text models, steered with sketches, references and masks. The prompt is one input among many.

A tango takes two. You lead, the AI follows.

or scroll down App details

Line drawing of an open kitchen and living room

The room after guided generation with the image kept from Explore

The same room with an orange and mint reference

The same room with a dark, marble and walnut reference

The same room with a walnut and white marble reference

The reference: the image kept from Explore

Upload a sketch.

Or upload a screenshot.

Then drop in the image you kept, as the reference.

Another reference, and the image transforms again.

Next. Edit.

Open the image with the mask editor.

Mask

Paint over what should change and say what you want: “add a woman sitting on the chair”.

Scribble

Draw where something should go: “replace the red area with a man sitting on the barstool while reading a newspaper”.

Scribble again

A new scribble changes the pose.

Reference material

Mask the floor and load a material: “replace the floor material with the grey stone material as seen in the reference image”.

Erase

Mask what should go: “get rid of the hanging pendant lamps”.

Whole image

Without a mask, the prompt changes everything.

Next. Styletransfer.

The same image in another style.

Styletransfer

The room stays. The style comes from a reference image.

Postprocess

Opened from the fast buttons.

Postprocess

Preset “Document White”, saturation down, exposure up: the tint becomes white paper.

Postprocess

Wipe compare: before and after.

Next. Upscale.

More resolution.

Upscale

The image comparer opens from the fast buttons.

Upscale

Before and after, zoomed in. Then the whole image again.

Explore to first result

A prompt, a sketch, a model, the first room.

Prototype

Three more references, three more rooms.

Edit

One image, six changes.

Style

One room, three styles.

The same room as a flat illustration in warm colours

The same room as a flat illustration in teal

The same room as a pencil sketch, as it comes out of the style transfer

This run is an interior design example.

It works exactly the same for architecture, content and industrial design.

This was a typical interior design run.

Here is an overview of its steps.

also orient yourself based on attached image.

You

The attached reference image: a living room with a marble fireplace

Local prompt assistant

Workflows

A custom, open library for dedicated use cases.

Fast buttons

Postprocess, AI Transparency, Latest outputs, Local Prompt Assistant, Image Comparer, Free GPU.

Workflow settings

Options to steer the generation.

Generation canvas

Where every result appears.

First, let’s start with a prompt.

Next. Guided generation.

Boards

A project needs more than one image. Further workflows turn the result into boards: a moodboard with materials and palette, an art direction board, and a reshoot board with new views and details of the same room. For the team to work with, and ready to show the client.

Moodboard

taupe#9d908525.2% taupe pale#b5aca622.4% taupe light#cdc8c516.6% walnut#7c695a15.5% umber#4d392d14.1% linen#eeeeef6.3%

Art Direction Board

Reshoot Board

Further use cases

For your office

Everyone in the office can work with it, without extensive training.
It works with 2D images: sketches, screenshots, photos. So it works with every common CAD program, no direct integration needed.
An open workflow library. It starts with a base set and grows with specialised workflows built only for your office: workflows no one else has, so the office keeps its own identity.
The know-how stays in the office, not with one person.
Your project data stays on your own hardware by default.
It is not only a visualisation tool: use it early, to find ideas and test directions.
Much of the visualisation work can be done in-house.

How it starts

You start with six workflows and choose two more.

We set Tango up in one day and train your team.

How it runs

Tango runs mostly in your office. A few steps go to an external model, where the quality calls for it. How much stays local depends on your machine and on how complex the task is. You decide, per project.

Mostly local: Recommended. Generation runs on your machine and sensitive steps stay in-house; quality-critical steps switch to an API.
Fully local: No external services. Some steps, certain edits for example, lose quality.
API only: No new machine. Tango runs on the hardware you have; all generation goes through external services, paid per call.

Specs

Hardware: Recommended: a dedicated Windows machine with NVIDIA RTX 4080 or better, 64 GB RAM, fast NVMe SSD. Mac possible.
Access: In the browser, from every workstation in the office, Mac or Windows
Data: Local by default. External services only where you switch them on.
Built on: ComfyUI
Models: Local models licensed for commercial work
Built in: Local prompt assistant, mask editor, postprocess, image comparer, image archive
Workflows: Six to start, two more of your choice
AI Transparency: Labelling under the EU AI Act
Price: On request

Who is behind it

A.I.gency is a visual communication studio in Vienna. It was founded by Simon Hirtz, who builds Tango himself and looks after every installation personally.

Contact

Let’s talk

Every office works differently. Write to us and we will arrange a call to see what makes sense for yours.

sh@a-i-gency.com Get in touch