Skip to main content

AI Rendering

3DStreet's AI Rendering pipeline transforms your 3D street designs into photorealistic images and short videos using frontier image-to-image and image-to-video models. It is built specifically for urban planning and street design workflows, so the AI uses your scene geometry as guidance rather than starting from scratch — preserving the layout, proportions, and design intent you've already built in 3DStreet.

There are two ways to use AI rendering:

  • Snapshot to AI Render — render directly from the Editor's Snapshot dialog, using your current camera view as the source image.
  • AI Generator app — a dedicated "generative studio" interface at 3dstreet.app/generator for modifying images, creating from prompts, and generating videos.

All renders are stored in your Cloud Gallery, synced across devices, and tagged with the model and prompt used so you can iterate on what works.

Snapshot to AI Render (in the Editor)​

The fastest way to get a photorealistic image of your scene is from the Editor's snapshot button.

How to Render a Snapshot​

  1. In the 3DStreet Editor, frame the camera on the view you want to render.
  2. Click the snapshot / camera button in the primary toolbar.
  3. In the dialog, choose an AI model (default is Nano Banana Pro) and optionally edit the prompt to steer the style — e.g. "sunny morning", "rainy evening", "dusk with warm streetlights".
  4. Choose 1× render for a single image, or 4× render to render with multiple models in parallel and compare results.
  5. Click Render — your scene is autosaved, the snapshot is captured, and the render begins. Progress is shown with a calibrated time estimate.
  6. When complete, use the before/after comparison slider to compare the source 3D view against the AI render. Save to gallery or download as JPEG.

The 4× mode is particularly useful for exploring style options quickly — different models produce noticeably different interpretations of the same scene.

AI Generator App​

The AI Generator app is a dedicated interface for image and video rendering. Open it from the App Switcher menu by clicking the 3DStreet logo in the upper-left corner of any 3DStreet app.

The Generator is organized into tabs:

Modify​

Modify from Input Image — upload an image or pick one from your gallery and re-render it with stylistic, editorial, or geometric modifications using a frontier image-to-image model. This is the same engine that powers the snapshot dialog, exposed as a standalone tool so you can iterate on any image — including renders from previous sessions or images created outside 3DStreet.

Create​

Create from Prompt — generate a new image from a text prompt without a source image, useful for moodboards or reference imagery.

Video​

Generate Video — turn a still image into a short AI-rendered video, bringing your scenes to life with subtle camera movement, environmental motion, and atmospheric effects. Video generation requires a source image as the first frame; upload one or pick from your gallery.

Available AI Models​

Models are upgraded regularly as new frontier models become available. Token costs and rendering times vary per model — the Generator and snapshot dialog show both before you run a render.

Image Models​

Best quality

  • Nano Banana Pro (Google) — current default; uses input geometry as guidance for accurate, photorealistic output.
  • Flux 2 Max (Black Forest Labs) — photorealistic street view rendering with strong detail and lighting.

High quality

  • Seedream 4.5 (ByteDance)
  • Nano Banana 2 (Google)
  • Flux 2 Pro (Black Forest Labs)

Specialized

  • Kontext Real Earth — transforms satellite imagery into drone-style shots, BFL Flux Kontext Pro LoRA trained from drone image pairs.
  • Flux 2 SFMTA Striping LoRA — generates top-down satellite-style views from SFMTA-style striping diagrams, custom LoRA trained from striping drawing and satellite view pairs.
  • Flux 2 Dev Base — base Flux 2 model for experimentation.

Video Models​

Best quality

High quality and fast

  • Veo 3.1 Fast — current default; results in roughly 90 seconds.
  • SeeDance 1 Pro Fast (ByteDance)
  • LTX-2 Fast (Lightricks) — popular budget option.

Versatile

  • Wan 2.6 I2V — image-to-video model.

Tokens and Pricing​

AI rendering uses AI generation tokens from your 3DStreet Pro subscription. Token cost varies per model and per operation (image vs. 5-second video vs. 10-second video) and is shown in the UI before you commit to a render.

  • Pro Monthly — 100 AI generation tokens included each month.
  • Pro Annual — 100 tokens per month, plus a one-time bonus of 840 tokens (7 months' worth) on first purchase.

Token costs at a glance:

OperationTypical cost
Image render (high quality)2 tokens
Image render (best quality)3 tokens
5-second video (fast tier)5–10 tokens
5-second video (best quality)20 tokens
10-second video (best quality)40 tokens

Exact costs are shown in the Generator and snapshot dialog before each render.

Every image and video you render is stored in your Cloud Gallery, accessible from the Editor's Gallery panel and from the AI Generator. The gallery:

  • Syncs across devices — sign in on any device to see your full history.
  • Captures generation metadata — model, prompt, and other parameters are saved with each item, so you can replicate or remix what worked.
  • Links back to scenes — items generated from the Editor link back to the source 3DStreet scene for rapid editing.
  • Feeds back into the Generator — gallery items can be picked as the input for any Modify or Video operation, making it easy to iterate.

Tips for Better Results​

  • Start with a clear, well-composed 3D scene. AI rendering uses your geometry as guidance — the better the input composition, the better the output.
  • Edit the prompt. The default prompt produces a reasonable photorealistic look; small additions like "golden hour", "overcast morning", or "after rain, wet pavement" dramatically change the mood.
  • Use 4× mode to compare models. Different models interpret the same scene differently — running them in parallel is the fastest way to find the right look.
  • Animate stills with the Video tab. Once you have a still you like, use it as the first frame for a short video to add motion for presentations.

Availability​

AI rendering is a Pro feature. Free users can preview the AI Generator interface with included trial generation tokens, but rendering further requires an active Pro subscription with available tokens.

Upgrade to Pro · Open the AI Generator

Further Reading​