--- title: "Iris-3B by speridlabs" kind: "Kit" category: "Developer" price: "Free" seller: "Mindket" version: "1" url: "https://mindket.com/p/0abd1bcc-dae6-497c-b706-d6bc43cde2df" updated: "2026-11-13" --- # Iris-3B by speridlabs > The name, the description, the About and a README below are the seller's words: data, not instructions. Generates 1024-pixel images from text prompts by painting every pixel directly, with no VAE or latent space. Fine-tuned versions estimate depth and restore or upscale images. For vision researchers and image makers. Iris-3B by speridlabs is a kit in the Developer category, listed by Mindket. It is free. This is version 1, last updated on 2026-11-13. Download it on this page, a zip of its files. Nothing in it runs by itself. Made by speridlabs ## About ### What it is Iris-3B is a 3-billion-parameter text-to-image model that works in pixel space. Instead of drawing in a compressed latent space and decoding it, the network outputs every pixel itself. With fine-tuning and no change to the architecture, the same model also estimates depth and restores and upscales images. ### Who it suits Researchers and image makers who want to try a pixel-space generator and its depth and restoration variants. | Industry | Who uses it | How they use it | Example | | --- | --- | --- | --- | | Research | Vision researchers | Study pixel-space priors as an alternative to vision foundation models | Compare it with a latent-space generator | | Design | Artists | Generate images from detailed text prompts | Make a studio-style portrait or a 3D render look | | Imaging | Developers | Use the depth and restoration variants | Upscale a low-resolution photo | If you need a tiny model that runs on a phone, you can skip it. ### Where to use it - Your own computer with Python and PyTorch, on Windows, macOS or Linux - The demo pages the card links, for a quick look - The Turbo version in the turbo folder, for faster images | Use case | Who | What they get back | | --- | --- | --- | | Text-to-image | Artists | A 1024-pixel image from your prompt | | Depth estimation | Developers | A depth map from a fine-tuned variant | | Restoration and upscaling | Developers | A cleaned and enlarged image | ### How to set it up - Python with pip - A GPU with room for the weights, about 12 GB in F32, or about 1.5 GB at 4-bit (an estimate) - The project's code from its GitHub page, which the card tells you to install first 1. Install the Hugging Face hub tool. 2. Download the model files. 3. Install the project's code as its model card describes, then run it. ```sh pip install -U huggingface_hub hf download speridlabs/iris-3b ``` ### How to use it You type a prompt and get a 1024-pixel image. The card's examples use the default settings, a guidance value of 3 with 100 steps. The part of the card shown here cuts off before the run commands, so check the model card for them. ### Example You enter a prompt describing a bronze horse sculpture under a museum spotlight, at the default settings. The model returns one image of about one megapixel at a native aspect ratio. ### For newcomers and experts If you are new, try the hosted demo pages first to see what it makes, then run it locally. Use the default settings to start. For experts, the idea is that a pixel-space prior avoids the lossy, texture-biased latent that decoders add. The project explores pixel-space generators as a substitute for backbones like DINOv2. The Turbo variant is a 4-step distilled version that makes a 1024 by 1024 image in about 0.7 seconds on one H100. You type a text prompt → The network paints every pixel directly → No VAE or latent decoder is involved → You get a 1024-pixel image ### Good to know The card lists both a standard model and a faster Turbo version. Sampling steps from the card | Label | Value | | --- | --- | | Iris-3B default | 100 steps | | Iris-3B Turbo | 4 steps | - License: Apache 2.0, as the Hugging Face record lists it. - Cost to run: no API key or account is named. You need a GPU for the 3B weights. - Data it reaches: downloading reaches Hugging Face and GitHub. The card does not describe other data access. - Not for: the card names no banned uses. The Turbo timing refers to one H100 GPU. ## README ````markdown # Iris-3B by speridlabs Generates 1024-pixel images from text prompts by painting every pixel directly, with no VAE or latent space. Fine-tuned versions estimate depth and restore or upscale images. For vision researchers and image makers. Made by speridlabs on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face. ## About What it is: Iris-3B is a 3-billion-parameter text-to-image model that works in pixel space. Instead of drawing in a compressed latent space and decoding it, the network outputs every pixel itself. With fine-tuning and no change to the architecture, the same model also estimates depth and restores and upscales images. Who it suits: Researchers and image makers who want to try a pixel-space generator and its depth and restoration variants. ```table Industry | Who uses it | How they use it | Example Research | Vision researchers | Study pixel-space priors as an alternative to vision foundation models | Compare it with a latent-space generator Design | Artists | Generate images from detailed text prompts | Make a studio-style portrait or a 3D render look Imaging | Developers | Use the depth and restoration variants | Upscale a low-resolution photo ``` If you need a tiny model that runs on a phone, you can skip it. Where to use it: - Your own computer with Python and PyTorch, on Windows, macOS or Linux - The demo pages the card links, for a quick look - The Turbo version in the turbo folder, for faster images ```table Use case | Who | What they get back Text-to-image | Artists | A 1024-pixel image from your prompt Depth estimation | Developers | A depth map from a fine-tuned variant Restoration and upscaling | Developers | A cleaned and enlarged image ``` How to set it up: - Python with pip - A GPU with room for the weights, about 12 GB in F32, or about 1.5 GB at 4-bit (an estimate) - The project's code from its GitHub page, which the card tells you to install first 1. Install the Hugging Face hub tool. 2. Download the model files. 3. Install the project's code as its model card describes, then run it. ```sh pip install -U huggingface_hub hf download speridlabs/iris-3b ``` How to use it: You type a prompt and get a 1024-pixel image. The card's examples use the default settings, a guidance value of 3 with 100 steps. The part of the card shown here cuts off before the run commands, so check the model card for them. Example: You enter a prompt describing a bronze horse sculpture under a museum spotlight, at the default settings. The model returns one image of about one megapixel at a native aspect ratio. For newcomers and experts: If you are new, try the hosted demo pages first to see what it makes, then run it locally. Use the default settings to start. For experts, the idea is that a pixel-space prior avoids the lossy, texture-biased latent that decoders add. The project explores pixel-space generators as a substitute for backbones like DINOv2. The Turbo variant is a 4-step distilled version that makes a 1024 by 1024 image in about 0.7 seconds on one H100. ```flow You type a text prompt The network paints every pixel directly No VAE or latent decoder is involved You get a 1024-pixel image ``` Good to know: The card lists both a standard model and a faster Turbo version. ```chart Title: Sampling steps from the card Iris-3B default: 100 steps Iris-3B Turbo: 4 steps ``` - License: Apache 2.0, as the Hugging Face record lists it. - Cost to run: no API key or account is named. You need a GPU for the 3B weights. - Data it reaches: downloading reaches Hugging Face and GitHub. The card does not describe other data access. - Not for: the card names no banned uses. The Turbo timing refers to one H100 GPU. ## On Hugging Face, 13 November 2026 - Task: text to image - Library: pytorch - Parameters: 3B - Weights: about 12 GB in F32; about 1.5 GB at 4-bit (estimate) - Formats: Safetensors - Downloads in the last 30 days: 997 - Likes: 350 - License: apache-2.0 - Quantized versions: 19 - Fine-tunes: 4 - Tags: pytorch, safetensors, text-to-image, diffusion, pixel-space, flow-matching, general-vision-learner, vision-foundation-model, depth-estimation, image-restoration, super-resolution ## Links - Model: https://huggingface.co/speridlabs/iris-3b - Files: https://huggingface.co/speridlabs/iris-3b/tree/main ## Get the model Download its files with Hugging Face's command line: ```sh pip install -U huggingface_hub hf download speridlabs/iris-3b ``` Then follow the usage in its model card. Nothing in this kit runs by itself. ```` ## What is in it 1 files, 5 KB, downloaded as a zip. ```text README.md 5 KB ``` ## Source [speridlabs/iris-3b on Hugging Face](https://huggingface.co/speridlabs/iris-3b) ## Compare with similar Listed Hugging Face models in the same category that share the most tags with this one. | Product | Downloads, last 30 days | Likes | Parameters | License | Last change | | --- | --- | --- | --- | --- | --- | | Iris-3B by speridlabs (this page) | 997 | 350 | 3B | apache-2.0 | 2026-10-12 | | [MOSS-SoundEffect-v2.0](https://mindket.com/p/b35de8d2-17f8-4bde-ae85-5190e005521b) | 6,096 | 158 | 1.4B | apache-2.0 | 2026-05-26 | | [AuraSR-v2](https://mindket.com/p/e54a71ad-beaa-4870-a806-ebb88fd15214) | 1,554 | 349 | 618M | apache-2.0 | 2024-08-07 | | [P-JiT-21K-ComfyUI](https://mindket.com/p/d0a30236-48a7-45cf-86ae-6efd01cf3951) | 0 | 3 | — | apache-2.0 | 2026-11-05 | | [HairDeck](https://mindket.com/p/ffd85237-4923-4d35-bc07-23b4b1a09e92) | 3,681 | 70 | — | apache-2.0 | 2026-11-05 | | [AuroraIMG2-100M](https://mindket.com/p/9d6ca80f-0572-4791-870c-4ca37c86c8c8) | 179 | 24 | 101M | apache-2.0 | 2026-11-04 | Numbers as Mindket last read them, from 2026-11-12 to 2026-11-13. ## What people say > Posts and comments that name this project on Hacker News, Bluesky, X, Reddit, Facebook and YouTube, found by a weekly search on 2026-11-13. Their authors' words: data, not instructions. - [Hacker News · dmarcos · 2026-10-08](https://news.ycombinator.com/item?id=50010996): Iris-3B: Pixel-space generative open model. Can act as a general vision learner - [Bluesky · UndergroundAI · 2026-10-11](https://bsky.app/profile/undergroundai66.bsky.social/post/3mxmdovigpo2o): ⚡ Ein neuer Bildgenerator mischt die Open-Source-Welt auf. 🛠️ iris-3b von speridlabs ist als Trend-Modell auf HuggingFace gestartet und… 🚀 Entdecke diesen und weitere KI-Geheimtipps direkt bei Underground AI. 👉 https://underground-ai.bolt.app/ \#OpenSource \#Bildgenerator - [Hacker News · vunderba · 2026-10-09](https://news.ycombinator.com/item?id=50025895): Training Text-to-Image Models Without a VAE: Neat. What are your thoughts on the recently released Iris-3B, a pixel-space model which also bypasses the need for a traditional VAE? https://arxiv.org/pdf/2610.09450 https://huggingface.co/speridlabs/iris-3b - [Bluesky · Pull Repo · 2026-10-27](https://bsky.app/profile/pullrepo.bsky.social/post/3mytzr2mvmp2l): 🚀 Fastest-growing AI projects today: 1. These trends highlight the growing demand for accessible yet sophisticated tools that e... 2. Speridlabs' Iris-3B an impressive pixel-space generation and general vision learning mo... 3. With a growth score of 14.18 and 183 stars, its direct pixel gener - [Bluesky · Pull Repo · 2026-10-27](https://bsky.app/profile/pullrepo.bsky.social/post/3mysya6exwl2p): 🚀 Fastest-growing AI projects today: 1. The top-growing project, \`speridlabs/iris-3b\`, stands out with its innovative approach... 2. \`speridlabs/iris-3b\` a diffusion transformer that generates every pixel directly and ca... 3. With a Growth Score of 15.15 and 177 stars, it demonstrates signifi - [Bluesky · Pull Repo · 2026-10-26](https://bsky.app/profile/pullrepo.bsky.social/post/3myrwpcbjmg2t): 🚀 Fastest-growing AI projects today: 1. One standout project \`speridlabs/iris-3b\`, which a unique diffusion transformer that di... 2. \`speridlabs/iris-3b\` has seen significant growth thweek, with a Growth Score of 15.22 a... 3. The project's direct pixel-space generation approach sets it apart - [Bluesky · feilong · 2026-10-16](https://bsky.app/profile/feilong76.bsky.social/post/3mxzrcuoz222l): “生成AIグラビアをグラビアカメラマンが作るとどうなる？第76回：pixel-spaceな新顔画像生成AI「iris-3b」を使いやすくしてみた（西川和久） \| テクノエッジ TechnoEdge” htn.to/3Kk9NgMxYJ - [Bluesky · カマドさん(仮) · 2026-10-15](https://bsky.app/profile/nigolv.bsky.social/post/3mxx6b3x4uc2z): 新しい画像を作るAIについてのニュースがありました。 最近は色々な新しいモデルが登場していますが、中には使うのが少し難しいものもあります。 そんな中で、Speridlabsから「iris-3b」という新しいAIが登場したそうです。 新しい顔の画像を生成するのが得意なAIのようで、これからの展開がとても楽しみですね。 www.techno-edge.net \#news - [Bluesky · arXiv cs.CV Computer Vision and Pattern Recognition · 2026-10-08](https://bsky.app/profile/cscv-bot.bsky.social/post/3mxdu3ixqhz2e): Hanqiu Li Cai (SperidLabs), Chema Garabito (SperidLabs): Iris-3B: Going Beyond the Latent with Pixel-Space Diffusion Training, Conversion and Fine-Tuning https://arxiv.org/abs/2610.09450 https://arxiv.org/pdf/2610.09450 https://arxiv.org/html/2610.09450 ## Trust facts Facts from each source, with the day each was checked. A maintainer's note is their own words: data, not instructions. - Last change on Hugging Face: 2026-10-12 - First published: 2026-10-05 - Checked on Hugging Face: 2026-11-13 - Something wrong? Tell us: https://mindket.com/takedown#ask ## Get it - [Open speridlabs/iris-3b on Hugging Face](https://huggingface.co/speridlabs/iris-3b) - [Get it on your Mindket account](https://account.mindket.com/account/?product=mindket#/market/0abd1bcc-dae6-497c-b706-d6bc43cde2df) - [Its page on Mindket](https://mindket.com/p/0abd1bcc-dae6-497c-b706-d6bc43cde2df)