Engine Room.
We believe in radical transparency. RenderHut provides direct access to the world's most powerful foundation models. Here is the technical breakdown of our stack.
The Model Matrix
| Capability | Foundation Engine | Standard |
|---|---|---|
| Turbo Image | Flux-2 DevLightning-fast diffusion with high prompt adherence. | 1K (1024px) |
| Smart Image | GPT-Image 1.5Excels at spatial logic and complex text rendering. | 1.5K (1536px) |
| Ultra Image | Seedream V5Native 2K photorealism with deep prompt reasoning and multi-element composition. | 2K (2048px) |
| HD Video | Kling v2.6 / Veo 3.1Cinematic temporal consistency and motion physics. | 1080p Native |
| Upscale | SeedVR 2.0Generative restoration and intelligent detail synthesis. | Up to 8K |
Deep Prompt Reasoning
The **Ultra Tier** runs on ByteDance's Seedream V5, a reasoning-guided image model. Instead of matching keywords to visual patterns, it interprets creative intent holistically — composition, lighting, spatial relationships — before rendering native 2K assets in a single pass.
Why the credit markup?
Your credit cost covers more than just raw GPU time. It includes the execution of our **Logic Engine** (Prompt Expansion), the **Multimodal Safety Guard**, and the **Vault Pipeline** (High-speed R2 ingress and master file archival).
Version Pinning
We periodically update the underlying foundation models. When a provider releases a new version (e.g., v2.6 to v3.0), we perform an internal A/B quality test before pinning the new version to our production studio.