tensorart/stable-diffusion-3.5-large-TurboX
68201
1---2license: other3license_name: stabilityai-ai-community4license_link: LICENSE5---6 7# TensorArt-TurboX-SD3.5Large8<div align="center">9<img src="ckpt_ad.webp" alt="ComfyUI_temp_psqad_00205_" width="400"/>10</div> 11 12## Model Description13 14**TensorArt-TurboX-SD3.5Large** is a highly optimized variant of **Stable Diffusion 3.5 Large**, achieving **6x faster generation speed** with minimal quality loss. It surpasses the official **Stable Diffusion 3.5 Large Turbo** in **image detail, diversity, richness, and realism**.15 16The model is available in both **LoRA** and **checkpoint (ckpt)** formats(chosen one of them, don't use them at same time), ensuring **compatibility with most community models**. This allows seamless integration into various workflows, accelerating **open-source AI development**.17 18## Key Features19 20- 🚀 **6x Speed Boost**: Generates images significantly **faster** than the original SD3.5 Large.21- 🎨 **Superior Quality**: Outperforms **stabilityai/stable-diffusion-3.5-large-turbo** in **detail, diversity, and realism**.22- 🔄 **Versatile Compatibility**: Works with both **realistic** and **anime-style** models.23- ⚡ **Optimized for Efficiency**: Suitable for both **high-end and mid-range GPUs**.24 25## Recommended Settings26 27For optimal results, we recommend using:28 29- **Sampler**: `Euler`30- **Scheduler**: `Simple`31- **Sampling Steps**: `8`32- **CFG Scale**: `1.0 - 1.5` _(Note: CFG ≠ 1 will double generation time)_33- **Shift**: `5` _(Most important！！！)_34- **LoRA Strength**: `1.0` _(if using LoRA version)_35## Example Comparisons36 37### TensorArt-TurboX-SD3.5Large vs. Official Models38 39Here are some sample outputs comparing **TensorArt-TurboX-SD3.5Large** (8 steps, CFG=1, Simple scheduler) with **official stable-diffusion-3.5-large-turbo** when using the same time. The contrast workflow are here: [constrast_normal_TurboX](./contrast_sd3.5L_normal_TurboX.json), [contrast_TurboX_turbo](./contrast_sd3.5L_TurboX_turbo_diff_cfg.json):40 41| prompt | Tensorart turboX cfg=1 8steps | Official turbo cfg=1.5 4steps |42| --- | --- | --- |43| | **scheduler=simple** | **scheduler=normal** |44| a high school girl is standing in the ground,full body, wearing a sport clothes, holding a tennis |  |  |45| Cinematic close-up on Fan Bingbing's regal features as she embodies ancient Egyptian queen Cleopatra. Golden light bathes her porcelain skin, highlighting the delicate contours of her face and the piercing gaze that commands attention. Rich fabrics drape elegantly across her statuesque form, with a subtle emphasis on the ornate jewelry and intricately designed headdress that crowns her majestic presence |  |  |46| A blonde woman in a short dress stands on a balcony, teasing smile and biting her lip. Twilight casts a warm glow, (anime-style:1.2). Behind her, a jungle teems with life, tropical storm clouds gathering, lightning flickering in the distance. |  |  |47 48## Example Output49 Using ckpt： 50 51 ```python52 import torch53 from diffusers import StableDiffusion3Pipeline54 55 repo = "tensorart/stable-diffusion-3.5-large-TurboX"56 57 58 pipe = StableDiffusion3Pipeline.from_pretrained(repo, torch_dtype=torch.float16)59 60 pipe = pipe.to("cuda")61 62 63 generator = torch.Generator(device="cuda").manual_seed(1)64 65 image = pipe(66 "A beautiful bald girl with silver and white futuristic metal face jewelry, her full body made of intricately carved liquid glass in the style of Tadashi, the complexity master of cyberpunk, in the style of James Jean and Peter Mohrbacher. This concept design is trending on Artstation, with sharp focus, studio-quality photography, and highly detailed, intricate details.",67 num_inference_steps=8,68 guidance_scale=1.5,69 height=1024,70 width=768,71 generator=generator72 ).images[0]73 74 image.save("./test2-3.webp")75 76 ```77 78 Using lora：79 80 ```python81 import torch82 from diffusers import StableDiffusion3Pipeline, FlowMatchEulerDiscreteScheduler83 import numpy as np84 from safetensors.torch import load_file85 from huggingface_hub import hf_hub_download86 87 repo = "tensorart/stable-diffusion-3.5-large-TurboX"88 ckpt = "Tensorart_TurboX_sd3.5L_8steps.safetensors"89 90 pipe = StableDiffusion3Pipeline.from_pretrained("stabilityai/stable-diffusion-3.5-large", torch_dtype=torch.float16,)91 92 pipe = pipe.to("cuda")93 pipe.load_lora_weights(hf_hub_download(repo, ckpt))94 pipe.fuse_lora()95 96 pipe.scheduler = FlowMatchEulerDiscreteScheduler.from_pretrained(repo, subfolder="scheduler", shift=5)97 98 generator = torch.Generator(device="cuda").manual_seed(1)99 image = pipe(100 "A beautiful bald girl with silver and white futuristic metal face jewelry, her full body made of intricately carved liquid glass in the style of Tadashi, the complexity master of cyberpunk, in the style of James Jean and Peter Mohrbacher. This concept design is trending on Artstation, with sharp focus, studio-quality photography, and highly detailed, intricate details.",101 num_inference_steps=8,102 guidance_scale=1.5,103 height=1024,104 width=768,105 generator=generator106 ).images[0]107 image.save("./test1.webp")108 ```109 110 111 112## Community Model Integration113 114**TensorArt-TurboX-SD3.5Large** seamlessly integrates with various **community models**:115 116- ✅ **Photorealistic Models**: Enhances realism while maintaining the original style.117- ✅ **Anime-Style Models**: Preserves artistic style while boosting generation speed.118- ✅ **LoRA Models**: Works with **style, character, and concept LoRAs**.119 120## Limitations121 122While highly optimized, this model may **not** be suitable for:123 1241. **Rendering precise English text** in images.1252. **Generating anatomically perfect hands**.1263. **Experiments with non-recommended samplers or schedulers** _(Best performance with Euler + Simple scheduler)_.127 128## Model Links129 130- **TensorArt (LoRA Version)**: [TensorArt-TurboX-SD3.5Large-8steps](https://tensor.art/models/834787530596537105/TensorArt-TurboX-SD3.5Large-8steps)131 132 133## Contact134* Website: https://tensor.art https://tusiart.com135* Developed by: TensorArt136