Iris 3B text-to-image samples: 1360x1024 renders in ~1 minute on Blackwell
b4silio · reddit · 2026-10-09
Author b4silio shared user-generated samples from Iris 3B (pixelspace generation), responding to community requests.
- Generated via a vibecoded framework, 52-61 seconds per 1360x1024 image on a Blackwell 6000 Pro
- Examples include product-style bread photography, an origami paper motorcycle, expressionist portraiture, and a Makoto Shinkai-style meteor night sky
- Demonstrates the model's ability to follow long, highly detailed style prompts
More from Multimodal
- ETH's SpaceFlow: training-free locally controllable 3D generation from text and primitives — ethz · 2026-10-09
- 100-object test: AI now turns text into working 3D game objects — Astra most reliable, Opus 5.5 prettier — 141_1337 · 2026-10-09
- vlo v0.3.1 ships with a tutorial for AI video-making workflows using MiniMax — PxTicks · 2026-10-09
- Gossip Goblin's AI film Gods Don't Give Gifts hits theaters in December — TomLikesRobots · 2026-10-09
- GPT 2.5 Flare prompt template generates photos that look like candid working-photographer shots — techhalla · 2026-10-09
- GPT-Image x Fable collaboration yields stunning Iliad book 1-2 illustrations — aiamblichus · 2026-10-09