3D Environments for VR: Gaussian Splatting, PanoGen, Text-to-Scene

A VR application developer gets a task: create a realistic office interior for employee training. Manually modeling every detail takes weeks. Photogrammetry requires on-site visits. The alternative is neural network generation of 3D scenes from text descriptions or a few photos. We implement such pipelines turnkey — 8+ years of experience and over 50 projects in VR/AR. Contact us for a consultation on integrating neural network generation into your project.

The main challenge of neural methods is balancing quality and performance. Many models produce high quality but require minutes per frame. VR demands 72+ FPS. We solve this through automatic LOD optimization and INT8 quantization of neural networks, reducing memory consumption without quality loss.

How we generate 3D scenes from text

We take a text description (e.g., "an abandoned laboratory with fluorescent lamps") and run it through diffusion models — SceneScape or Set-the-Scene. The result is a depth map and semantic segmentation, which are converted into a mesh. An alternative route is PanoGen for 360° panoramas: 30–60 seconds per environment for initial prototyping. To enhance realism, we use few-shot fine-tuning on a small set of reference images.

Step-by-step scene generation process

1. Prepare a text description or set of photos (20–100 shots).
2. Choose the method: Gaussian Splatting for photos, PanoGen for 360°, Text-to-Scene for text.
3. Optimize: quantize the model to INT8, generate LODs, configure occlusion culling.
4. Export to the target engine (Unreal, Unity, WebXR).
5. Test on the headset (Quest 3, Pico) with FPS and memory metrics.

Why Gaussian Splatting outperforms NeRF for VR

NeRF delivers high detail but requires seconds per frame. Gaussian Splatting renders in real time without quality compromise. We use it for object reconstruction from 20–100 photos. Additionally, we apply INT8 quantization, reducing memory consumption by 2–3 times. Automatic LOD generation using a quadric error metrics simplification algorithm achieves 72 FPS even on mobile headsets. We guarantee compatibility with any target device after tuning.

Typical errors in 3D scene generation and how to avoid them

When using Gaussian Splatting, artifacts often appear at object boundaries. The solution is to add depth regularization and tune the number of iterations. Text-to-Scene may produce unrealistic proportions — we fix this by adding semantic maps and fine-tuning on an interior dataset. All our pipelines include automatic quality control with PSNR/SSIM metrics. Time savings compared to manual modeling reach 80%, and generation costs are 5–10 times lower.

When does generative 3D generation replace classic modeling?

For mass production of variations of the same environment — for example, 50 different offices for negotiation training — generative methods are indispensable. When time is tight, a text description turns into a draft scene in minutes. For unique high-detail assets, manual modeling remains the primary option. Budget savings on mass generation reach 60% compared to manual work. Contact us for a preliminary assessment of your project — we will select the optimal generation method for your budget and timeline.

What is included in the work

Analysis and design

Audit of input data, selection of architecture (text/photo/geometry). Choice of methods (Gaussian Splatting, NeRF, PanoGen) and pipeline configuration based on the target device.

Implementation and integration

Training and quantization of models, automatic VR optimization. Export to Unreal Engine, Unity, WebXR. Plugin development for your project. All stages are documented.

Testing and support

Performance measurements on target devices (Quest 3, Pico, etc.) with FPS and memory metrics. Training your team, documentation, and one month of support.

Performance metrics (example on Quest 3)

Final scenes are exported in formats: UAsset (Unreal Engine 5), Prefab (Unity), glTF 2.0 (WebXR), OpenXR/GL (Quest 3). Our certified specialists guarantee correct operation on any headset. Request an assessment of your project — we will select the optimal generation method for your budget and timeline.

Generative AI Development: From Prompt to Production API

We often receive a task "generate a product image" — on the surface it seems simple. But behind this lies a choice between dozens of models, configuring the inference pipeline, manually solving consistency issues, integrating into the product backend, and answering why the model generates hands with six fingers in staging but not in production. Let's break down the directions we work with.

Image Generation: From Prompt to Production API

The current landscape includes FLUX.1 [dev/schnell/pro] from Black Forest Labs and Stable Diffusion 3.5. FLUX.1 [schnell] takes 4 steps instead of 20–50 for SDXL — 5–12 times faster — while maintaining higher quality. On an A100 80GB — 1.2–1.8 s per 1024×1024 image at batch_size=4.

A typical deployment issue: FLUX.1 [dev] requires 24+ GB VRAM in fp16. On A10G 24GB it fits tightly; at batch_size>1 — OOM. Solution: torch_dtype=torch.bfloat16 + enable_model_cpu_offload() from diffusers, or quantization via bitsandbytes to NF4 — minimal quality drop, memory consumption drops to 12–14 GB.

ControlNet and IP-Adapter are key tools for production tasks where controllability is needed. ControlNet with Canny/Depth/Pose maps provides structural control. IP-Adapter (especially IP-Adapter-FaceID) allows transferring character identity to generations — this is the foundation for personalized content. More about ControlNet can be found on Wikipedia.

Case study: e-commerce photography. A retailer with 8000 SKUs needed lifestyle photos for each product. Pipeline: product segmentation (Segment Anything Model 2) → background removal → inpainting with FLUX.1 [dev] using product image as IP-Adapter reference → upscale via RealESRGAN_x4plus. The generation cost is negligible compared to professional photography, providing huge savings. Throughput — 200 images/hour on 2× A100. Our extensive experience from 30+ projects ensures we select the optimal model for your task — an evaluation can be obtained upfront.

Why Is Model Selection Only Half the Battle?

Fine-tuning for a Specific Style or Character

Dreambooth and LoRA are the standard for adapting to a specific visual style or object. LoRA trains in 2–4 hours on 20–30 reference images on a single A100. Rank 16–32 is usually sufficient for style; rank 64+ is needed for precise face reproduction.

A common mistake: training LoRA too long — the model overfits to references, losing the ability to vary. Sign: at cfg_scale=7, all images look like copy-paste of references. Solved by early stopping (usually 1500–2000 steps for 20 images) and prior_preservation_loss.

For deeper customization — full fine-tuning via diffusers + accelerate with FSDP on multiple GPUs. But that already takes 40–80 hours of training and requires a truly large dataset (1000+ images).

Comparison of Image Generation Approaches

Video Generation: Which Models Are Best?

Open-weight video models still lag behind commercial ones in stability and length. Wan2.1 is the best choice for self-hosting: 14B parameters, runs on 2× A100, delivers acceptable quality for short clips.

The main pain of video generation is temporal consistency: the character changes clothing color at the third second, objects "drift." Partial solution — generation with motion_bucket_id and noise_aug_strength in Stable Video Diffusion, or using I2V (image-to-video) instead of pure text-to-video. As noted in VideoPoet research, consistency is achieved by training on long sequences.

AnimateDiff remains a working tool for short loops and motion effects on top of SD/FLUX. Not Sora, but deployable locally and predictable.

Music and Audio Generation

AudioCraft from Meta (MusicGen + AudioGen) is a production-ready stack for music generation. musicgen-large (3.3B) generates 30 s of music in ~8 s on A100. Control via text prompt and melody conditioning — you can specify a melody by humming.

Stable Audio Open from Stability AI is an alternative with length up to 47 s, better structural control (intro/verse/chorus). Deployment is similar: diffusers + FastAPI.

For voice-over and dubbing — ElevenLabs API or self-hosted XTTS v2 (see Speech AI service). For sound design and foley — AudioGen.

3D Generation: Current Practical State

3D generation has not yet reached the same maturity as 2D. But for specific tasks, tools are already working:

TripoSG and Shap-E — text/image-to-3D. Shap-E from OpenAI generates simple 3D meshes in seconds, but geometry is rough. TripoSG gives more detailed results but requires post-processing (remeshing, UV unwrapping).

Wonder3D and Zero123++ — 3D reconstruction from a single image. They work by generating multi-views (6–8 views) and then 3D reconstruction via NeuS or instant-ngp.

Gaussian Splatting (3DGS) — not generation, but reconstruction from a series of photos/videos. For product cards and real estate it's already production: 50–200 photos → 3DGS model in 15–30 min on RTX 4090 → interactive 3D viewer in browser.

What Infrastructure Is Needed for Generative AI Deployment?

Critical for generative models:

• Task queue — Celery + Redis or Ray Serve. Synchronous HTTP for image generation is unacceptable with >5 concurrent requests.
• Caching — similar prompts yield similar results. Semantic cache via embeddings (faiss + sentence-transformers) can reduce GPU load by 20–40%.
• Quality monitoring — CLIP score for text-image alignment, FID for evaluating generation distribution. Integrate into MLflow or Weights & Biases.
• Storage — generated images immediately to S3/MinIO, not on the inference server disk.

What's Included in the Deliverables

We take the project turnkey — from model selection to deployment and monitoring. The result includes:

• Model (or API integration) with performance benchmarks (latency p99, throughput).
• Pipeline documentation (prompt engineering guide, model card, dependency versions).
• Integration with your backend (REST/gRPC, queues).
• Configured monitoring (dashboards, alerts for quality drift).
• Training workshop for the team (2–4 hours).
• Warranty support for 3 months after launch — as part of our quality certificate.

We have completed 30+ projects in generative AI — this gives us the right to guarantee results.

How Is the Generative AI Development Process Structured?

1. Analysis (1–2 days): audit of current architecture, clarification of use case, selection of models and success metrics. We evaluate the project free of charge.
2. Proof of Concept (1–3 weeks): quick prototype on your data — to see real quality, not blog demos.
3. Design (1–2 weeks): pipeline architecture, infrastructure (GPU cluster/API), A/B testing plan.
4. Implementation and fine-tuning (4–12 weeks): development, LoRA/full fine-tuning, integration with queue and cache.
5. Testing (1–2 weeks): load tests, metric validation, edge-case verification (negative scenarios).
6. Deployment and monitoring (1–2 weeks): production deployment, monitoring setup, documentation.

Timeline Estimates

Integration of a ready API (DALL·E 3, Midjourney API, Stability API) — 1–2 weeks. Self-hosted pipeline with fine-tuning — 6–12 weeks. Full platform with UI, queues and monitoring — 3–6 months. The specific cost is calculated individually after analyzing your scenario.

Contact us — order a consultation, and we will select the optimal architecture for your project. Get a preliminary cost and timeline estimate for free.