speridlabs/iris-3b · Hugging Face

Source: https://huggingface.co/speridlabs/iris-3b Pixel-space generation & general vision learner. A 3-billion-parameter model that paints every pixel directly — no VAE, no latent space — and, fine-tuned, estimates depth and restores and upscales images.

Generative priors are a promising foundation for downstream vision tasks. In this project we explore pixel-space generative models as an alternative to vision foundation models such as DINOv2.

Most image generators work in a compressed “latent” space and rely on a separate decoder to turn that into an image. Iris-3B skips that step: the network itself outputs every pixel, so nothing is lost to a lossy, texture-biased latent. You type a prompt, and you get a 1024-pixel image.

Iris-3B is not only an image generator. The same pixel-space prior, fine-tuned with no architectural change, is a general vision learner for dense tasks where detail matters: monocular depth estimation and image restoration and upscaling ship in this repo (seeA general vision learnerbelow).

https://huggingface.co/speridlabs/iris-3b#examplesExamples

Generated at native aspect ratios of about one megapixel with the default settings (CFG 3, 100 steps).

https://huggingface.co/speridlabs/iris-3b#get-startedGet started

1. Install the code

git clone https://github.com/speridlabs/iris-3b.git
cd iris-3b
pip install -e .          # Python 3.11+, PyTorch 2.7.1+

2. Download the weights(about 12 GB for text-to-image; thedepth/andupscaler/folders add about 12 GB each)

hf download speridlabs/iris-3b --local-dir iris-3b --exclude "depth/*" "upscaler/*"

3. Generate an image

python scripts/sample.py --checkpoint iris-3b \
    --prompt "a red fox sleeping in fresh snow, golden hour"

The text encoder (Qwen3-VL-4B-Instruct) downloads automatically on the first run. You need an NVIDIA GPU with CUDA.

https://huggingface.co/speridlabs/iris-3b#tipsTips

SettingDefaultWhat it does\-\-cfg\-scale3How strictly the image follows the prompt. Higher values follow the prompt more literally but can look harsher.\-\-steps100Number of denoising steps. Fewer steps are faster but lose some detail.\-\-seed—Fix it to get the same image again.\-\-negative\-prompt—Things you don’t want in the image.\-\-txt\-file—A text file with one prompt per line, for batches. Default output is 1024×1024. Write prompts as plain descriptive English sentences.

https://huggingface.co/speridlabs/iris-3b#a-general-vision-learner-depth-image-restoration-and-upscalingA general vision learner: depth, image restoration and upscaling

The same backbone, fine-tuned per task: Iris-3B estimates monocular depth and restores and upscales images. Both run the whole model in a single forward pass with the empty prompt, so no text encoder is needed. The weights download on first use.

Monocular depth— photo → relative depth:

Image restoration and upscaling— degraded input → restored image:

Try both interactively in thedemo, or from the command line:

python scripts/depth.py photo.jpg --out depth_out      # <name>.npy + colorized <name>.png
python scripts/upscale.py photo.jpg --out upscaled     # <name>_x4.png

TaskTrainingResultDepthDirect regression of relative log depth, 10K stepsAbsRel 0.071, δ1 0.946 (mean of NYUv2, KITTI, ETH3D, ScanNet, DIODE; Marigold V2 protocol, zero-shot)Image restoration and upscalingOne-step adversarial restoration (HYPIR recipe), 10K steps, EMA weightsLPIPS 0.292, PSNR 20.68 (DIV2K validation, 4×) Depth is affine-invariant, not metric: fit a scale and shift in log space to compare with ground truth. Restoration tiles large outputs into overlapping 1024×1024 windows.

https://huggingface.co/speridlabs/iris-3b#whats-in-this-repoWhat’s in this repo

FileContentsmodel\.safetensorsText-to-image weights (EMA, float32; run with bfloat16 autocast)config\.yamlArchitecture and sampling settings read byscripts/sample\.py``depth/Depth model:model\.safetensors,empty\_prompt\.safetensors,config\.yaml``upscaler/Image restoration and upscaling model: same three files

https://huggingface.co/speridlabs/iris-3b#about-the-modelAbout the model

Size3B parameters (plus a frozen 4B text encoder)ArchitectureDiffusion transformer: 8 dual-stream + 16 single-stream blocks, then a 4-block pixel head that turns each 16×16 patch back into pixelsText encoderQwen3-VL-4B-Instruct (frozen)TrainingRectified flow, trained from scratch at 256 → 512 → 1024 px, then supervised fine-tuning at 1024 px (665K steps total) The full training recipe is in thecode repository.

https://huggingface.co/speridlabs/iris-3b#limitationsLimitations

• Like every image generator, Iris-3B can produce inaccurate, biased or unsafe content. Review outputs before using them.
• Text rendering inside images and exact object counts are not always reliable.
• Images are generated at about one megapixel; for larger prints, use an upscaler.

https://huggingface.co/speridlabs/iris-3b#licenseLicense

Released under theApache License 2.0. The text encoder, Qwen3-VL-4B-Instruct, is downloaded from its publisher and stays under its own license (Apache 2.0).

https://huggingface.co/speridlabs/iris-3b#citationCitation

@article{licai2026iris,
  title   = {Iris-3B: Going Beyond the Latent with Pixel-Space
             Diffusion Training, Conversion and Fine-Tuning},
  author  = {Li Cai, Hanqiu and Garabito, Chema},
  journal = {arXiv preprint arXiv:2610.09450},
  year    = {2026},
  url     = {https://arxiv.org/abs/2610.09450}
}

Made with ❤️ bysperidlabs.com