Skip to content
Other

Iris-Z

a unified image and depth map generation model

2 min readupdated

Built on FLUX.2 by Speridlabs, Iris-Z is a generative model that simultaneously produces a visual image and its underlying geometry (depth map). Using a single set of weights, it handles text-to-RGBD, image-to-depth, and depth-to-image tasks. It outperforms leading dedicated depth estimators on standard benchmarks while preserving the generation quality of its base architecture, making it a strong fit for researchers, developers, and creators working in 3D graphics and computer vision.

Key features

  • Dual output: Generates both visual appearance and object geometry simultaneously from a single text prompt.
  • High accuracy: Beats top dedicated depth models on standard benchmarks like NYUv2 and ETH3D.
  • Flexible pipeline: Supports text-to-image, text-to-RGBD, image-to-depth, and depth-to-image workflows.
  • Two model sizes: Comes as the 32.2B-parameter Iris-Z-big and the lighter 9.1B-parameter Iris-Z-small.

What you can do

  • Text-to-RGBD: Generate a complete image alongside its matching depth map for 3D scenes straight from a text description.
  • Photo depth estimation: Extract detailed depth maps from an existing photo or batch-process an entire folder of images.
  • Depth-guided generation: Create new images guided by an existing depth map and a text prompt.
  • Python API integration: Plug the model directly into custom graphics pipelines using the official SDK.

Example

from iris_z import IrisZPipeline

pipeline = IrisZPipeline.from_pretrained("speridlabs/iris-z", variant="big")

# Генерация картинки и карты глубины по тексту
rgbd = pipeline.text_to_rgbd("a lighthouse at dusk", seed=0)

# Получение карты глубины из картинки
depth = pipeline.image_to_depth("photo.jpg")
depth.colourized().save("photo_depth.png")
Read moreDetails
AbsRel x100 benchmarks (lower is better): NYUv2 (Iris-Z-big: 1.90/2.28), ETH3D (Iris-Z-big: 1.76/1.87), depending on the evaluation protocol.
VRAM requirements: ~112 GB for the big version and ~33 GB for the small version across all tasks.
Requires accepting the base FLUX.2-dev model license and logging in via hf auth login.
Examples
Discuss Telegram 0

Will it run?

Runs in 4-bit

Full precision does not fit, a quantised version does: quality barely changes down to 8 bits, more noticeably at 4 bits.

  • BF16weights 60 GB · ready files66 GB
  • FP8 / INT8recommendedweights 30 GB33 GB
  • 4-bitweights 17 GB19 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # Infinix-io/iris-z
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download Infinix-io/iris-z --local-dir ~/models/iris-z
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to iris-z
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of FLUX.2-dev (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 12All on Hugging Face

About the model

Infinix-io/iris-z
Author
Infinix-io
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments