Shuttle3Diffusion：可商用F.1去蒸馏模型！仅4步约3秒出图，性能质量显著提升

🌹大家好！欢迎来到破狼公众号。感谢大家的支持与鼓励。在AIGC探索道路上，我将与你一路同行。喜欢就星标关注破狼公众号或文末扫码加入交流群 ！本人仅运营公众号平台，未经授权严禁CSDN等其他平台抄袭和转载！

‍

Shuttle 3 Diffusion简介

今天文章主题介绍一款最新的文生图模型：Shuttle 3 Diffusion 。这是一个使用Flux.1 Schnell作为基础的模型，是在仅通过4步骤根据文本提示语文生图创建AIGC图像的文生图模型。官方称：Shuttle 3 Diffusion模型在图像质量、排版、理解复杂提示和资源效率方面提供了增强的性能。同时提供了bfloat16、GGUF、fp8多个变体版本，方便不同显存用户使用。不同于Flux-dev模型 Shuttle 3 Diffusion是Apache 2开源许可，意味着可以用于商业用途使用。

另外，研发团队称：仅四步内就能产生与Flux Dev或Pro类似的图像。模型在训练过程中部分采用了Flux-Schnell去蒸馏方法。当使用超过10步时，它进入 “细化模式” ，增强图像细节而不改变构图。因此Shuttle 3 Diffusion采用了特殊的训练方法克服了Schnell系列模型的限制，从而实现了细节和颜色的表现的提高。

• • Shuttle 3 Diffusion BF-16：https://huggingface.co/shuttleai/shuttle-3-diffusion

• • Shuttle 3 Diffusion GGUF：https://huggingface.co/shuttleai/shuttle-3-diffusion-GGUF

• • Shuttle 3 Diffusion FP8：https://huggingface.co/shuttleai/shuttle-3-diffusion-fp8

• • 在线体验：https://chat.shuttleai.com/images

• • 模型介绍文档：https://docs.shuttleai.com/

演示案例

Shuttle 3 Diffusion ComfyUI体验

Shuttle 3 Diffusion是基于Flux-Schnell去蒸馏方法模型，因此 Shuttle 3 Diffusion的使用方式与Flux相同，仅需切换Unet加载模型即可。同时也可以使用在线体验。

• • 模型下载：https://huggingface.co/shuttleai/shuttle-3-diffusion/resolve/main/shuttle-3-diffusion.safetensors?download=true

• • 在线体验地址：https://chat.shuttleai.com/images

Shuttle 3 Diffusion在线体验

Shuttle 3 Diffusion文生图工作流

Flux文生图感兴趣的同学可参考LIBLIB在线运行工作流：FLUX[续篇]：12B参数23G最大开源文生图模型，Dev版直出惊艳美图欣赏

另外LIBLIBAI已支持本地客户端使用可首页下载体验。

注意：

• • 因为仅需4步加速，4张图896*1152分辨率仅需11.81秒加速，单图不到3秒完成。低显存友好，可商业使用。
• • 整体测试质量仅能说有提升但不及Flux.1-Dev，胜于Flux-Schnell原生模型有显著的质量提升，相比于dev模型则特别在于色彩、人物手部细节、背景处理方面不成熟。在风景等非人物细节绿具有较好质量。并且官方已有成熟商业化网站建设。另外，还可以增加lora和多次采样细节处理，提升整体出图质量。
• • Shuttle 3 Diffusion模型核心价值为：可企业商用并加强后期素材训练；4步加速针对低显存用户将是有利竞争。
  

01. 咖啡馆

Afternoon, Asian long-haired girl, sundress with a halter, Huaihai Road cafe, natural light, close-up, shallow depth of field, soft tones, elegant, relaxed, diagonal composition, high contrast, indoor, warm tones, 50mm lens, side lighting, slow shutter speed, smile, comfortable. Environmental light.,,in clothes

02.绿色印画

A green leaf with the outline of an young women drawn on it, held up against the blue sky through a glass window. The light is bright and sunny outside. There are visible water droplets in the foreground.

Flux-Dev	Shuttle 3 Diffusion

03.善与恶

This is a high-resolution,everyday scene image with a natural style,Create a movie poster with a dual-view design. On one side, depict a serene angelic figure against a pure white background, featuring a 20-year-old Chinese girl with an ethereal glow and delicate wings. On the other side, transform her into a captivating devil with striking horns and a fiery aura, set against a deep black background. Both images should seamlessly merge into one dynamic poster, capturing the contrast between light and dark, good and evil.

04.职业装

which is a high-resolution photograph featuring an Asian woman in a professional yet provocative office setting. The woman, likely in her mid-20s, has a petite physique with a slender build and fair skin tone. She is standing in an open doorway, with her left arm resting on the frame and her right hand holding a closed dark blue folder against her hip. She wears a crisp, white, long-sleeve blouse with a fitted cut that accentuates her small breasts, tucked into a tight black pencil skirt that clings to her slim waist and hips. Her outfit is complemented by sheer black pantyhose and black high heels with red soles. Her brown hair is styled in a neat, low ponytail, and she wears thin-framed glasses, adding a touch of sophistication. A black lanyard with a name badge hangs around her neck, and she has a confident, slightly sultry expression on her face. The background reveals a modest office setting with beige curtains, a wooden desk, and a red fire extinguisher placed discreetly in the corner. The overall scene exudes a mix of professionalism and allure, suggesting a provocative yet respectful workplace environment.

Flux-Dev	Shuttle 3 Diffusion

05.动漫风

Japanese anime style, Asian woman, 19 year old woman, she has an (oval shaped face:1.2), she has a (round chin:3.0) and a heavy lower lip, she has very long auburn hair, she has a (small upturned nose:1.3), large black framed glasses, thin lips, perfect skin, (hollow cheeks:1.2), perfect face, smiling, her wavy hair flows in front of one of her eyes . she's wearing dangling gold earrings, simple chain necklace with a pendant (baggy cropped cowl neck sweater:1.5), (large teardrop breasts:1.3), sweater off one shoulder coffee mug with latte art on table she's sitting in cafe, she's smiling at the viewer with an attentive expression leaning slightly forwards

Flux-Dev	Shuttle 3 Diffusion

06.小猫和鱼

A fluffy black cat is watching the small fish next to the fish tank, holding a fork, ready to poke the fish, in an exaggerated pose, against a white background, focusing on the cat. Simple background, close-up of the cat's upper body, plush, cute, funny, minimal illustration style, cartoonish chaos, nostalgic illustration, Amedee Ozenfant, Flickr, Miwa Komatsu, hand-drawn animation, high-definition details. 

Flux-Dev	Shuttle 3 Diffusion

更多推荐文章：

• Mochi1：更简单和可商用，ComfyUI内核支持，社区生态推动4张H100到消费级显卡可用

• CogVideo:重磅升级！图生视频完美镜头控制和3D环绕,商用级开源AI视频曙光

• 阿里InContextLoRA：更强ID一致性！基于黑森林F1身份一致性连贯视频分镜图集，10组风格无限创意

• Flux-NewReality:栩栩如生摄影级解禁模型,追求真实细节&风景&神话高品质艺术

• [ComfyUI]InstantIR：来自小红书团队模糊图像修复技术，效果是否惊艳？

• 更像了！5个百分点提升，字节写真换脸PuLID-F1再升级，小红书流量密码

• OmniGen：统一图像生成和多任务集成模型，任意人物自由合影，8位量化体验

• [ComfyUI]Flux：F.1多区域精确控图，无需LORA技术多区域自由构图工具

• [ComfyUI]MochiEdit：最新视频编辑工具，Mochi视频生成加速方案

• [ComfyUI]FaceAging：太好玩啦！仅需几秒看完你或她的一生，从出生到百岁面容

• [ComfyUI]Flux:低显存救星,无限创意!无需部署就能体验最新Joy2|PuLID|LLM等,CF无缝集成

感兴趣加入[AGI技术交流群]+V

如果觉得文章不错，就请赞、在看与转发三连