Shuttle 3 Diffusion - CivArchive (CivitAI Archive)

Shuttle 3 Diffusion - fp8

Shuttle 3 Diffusion

Shuttle 3 Diffusion is a text-to-image AI model designed to create detailed and diverse images from textual prompts. It offers enhanced performance in image quality, typography, understanding complex prompts, and resource efficiency.

You can try out the model through a website at https://chat.shuttleai.com/images

Using the model via API

You can use Shuttle 3 Diffusion via API through ShuttleAI

- ShuttleAI

- Documentation

Shuttle 3 Diffusion uses Flux.1 Schnell as its base. It can produce images similar to Flux Pro in just 4 steps, and it is licensed under Apache 2. The model was partially de-distilled during training. When used beyond 10 steps, it enters "refiner mode," enhancing image details without altering the composition. We overcame the limitations of the Schnell-series models by employing a special training method, resulting in improved details and colors.

Description

FAQ

Comments (32)

amazingbeautyNov 13, 2024

CivitAI

gguf?

raidmachine132017712Nov 13, 2024· 1 reaction

Yes please GGUF

robomaraiartNov 13, 2024· 1 reaction

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

zdwwNov 13, 2024

CivitAI

anime result looks good, waiting for fp8

_Envy_Nov 13, 2024· 1 reaction

https://huggingface.co/shuttleai/shuttle-3-diffusion-fp8/blob/main/shuttle-3-diffusion-fp8.safetensors

zdwwNov 13, 2024

@_Envy_ thanks

_Envy_Nov 13, 2024

@Kotflocke I'm not having any problems with making anime on it.

https://civitai.com/posts/9102007?returnUrl=%2Fmodels%2F943001

_Envy_Nov 13, 2024· 3 reactions

CivitAI

You should probably upload the fp8 and gguf versions here too. :)

shuttleai

Author

Nov 13, 2024

Thanks for the suggestion, uploading it right now!

RuoarNov 15, 2024

@shuttleai Thank you. :)

ericreatorNov 15, 2024

@Ruoar @shuttleai are there gguf somewhere?

ericreatorNov 15, 2024· 1 reaction

found it thx https://huggingface.co/shuttleai/shuttle-3-diffusion-GGUF

kunalpuriNov 13, 2024· 1 reaction

CivitAI

Great job on this checkpoint! It may not beat Dev but it does a lot better than the base Schnell.

arekksNov 14, 2024· 4 reactions

CivitAI

Great work here. Gens are clearly more detailed & appealing than base Schnell given the same step count & gen time. I'm also happy that the LoRAs I trained for base Schnell using Ostris's training adapter & ai-toolkit work here - in fact, the results are markedly better than the base model.

RalFingerNov 15, 2024

CivitAI

Thank you for the upload, great model!

yaode360276Nov 16, 2024· 1 reaction

CivitAI

is it better than flux1Lite8BAlpha?

raidmachine132017712Nov 16, 2024· 1 reaction

CivitAI

Need to work on better anatomy and realism, otherwise its go to model right now.

tigerartNov 17, 2024· 1 reaction

Yes. :( Plastic skin with no texture and inflamed red cheeks when trying to get photorealistic faces. You can see it everywhere in other users' examples.

shuttleai

Author

Nov 17, 2024· 3 reactions

Thanks for the feedback! A new version is in training, with more real photos.

RuoarNov 17, 2024

CivitAI

I like the composition and the prompt adherence is great but I don't like the wet oil paint look of it. I have been running images through slight upscaling (1.2x) and it cleans it up surprisingly well.

shuttleai

Author

Nov 17, 2024· 2 reactions

Thanks for the feedback! A new version is in training, with more real photos.

uprz23736Nov 18, 2024

CivitAI

Is the token limit 256 or 512?

dddimishNov 27, 2024

CivitAI

I like the images (especially the surreal ones) this model produces. I usually do the first two steps of 4 (or 6) to 1025x576, then two steps of scaling to 1920x1080. Works well and fast, but there is a problem... It sees figures and faces everywhere. ) Any weird noise turns into a person or a head. I even started using another model for scaling, and this one for generating the initial image and idea. I like to experiment with it.

mweldonsd594Dec 2, 2024· 1 reaction

CivitAI

This looks fantastic and I really want to give this model a spin, but I only have a 3060. So I would greatly appreciate it if you could post a 4 or 5-bit GGUF of this.

ericreatorDec 11, 2024

Ah that's tough. If you want a good gguf low vram workflow try flux Valhalla! https://civitai.com/models/818589/flux-valhalla

belemberDec 20, 2024· 1 reaction

CivitAI

This is an awesome model and it is the main model that I am using the most. Thank you and I am looking forward to updates.

KWFbDec 23, 2024

CivitAI

``` # ComfyUI Error Report ## Error Details - **Node ID:** 361 - **Node Type:** SamplerCustomAdvanced - **Exception Type:** RuntimeError - **Exception Message:** mat1 and mat2 shapes cannot be multiplied (693x768 and 4096x3072) ## Stack Trace ``` File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/execution.py", line 328, in execute output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/execution.py", line 203, in get_output_data return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/execution.py", line 174, in _map_node_over_list process_inputs(input_dict, i) File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/execution.py", line 163, in process_inputs results.append(getattr(obj, func)(**inputs)) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy_extras/nodes_custom_sampler.py", line 633, in sample samples = guider.sample(noise.generate_noise(latent), latent_image, sampler, sigmas, denoise_mask=noise_mask, callback=callback, disable_pbar=disable_pbar, seed=noise.seed) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/samplers.py", line 897, in sample output = executor.execute(noise, latent_image, sampler, sigmas, denoise_mask, callback, disable_pbar, seed) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/patcher_extension.py", line 110, in execute return self.original(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/samplers.py", line 866, in outer_sample output = self.inner_sample(noise, latent_image, device, sampler, sigmas, denoise_mask, callback, disable_pbar, seed) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/samplers.py", line 850, in inner_sample samples = executor.execute(self, sigmas, extra_args, callback, noise, latent_image, denoise_mask, disable_pbar) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/patcher_extension.py", line 110, in execute return self.original(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/samplers.py", line 707, in sample samples = self.sampler_function(model_k, noise, sigmas, extra_args=extra_args, callback=k_callback, disable=disable_pbar, **self.extra_options) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/pyt2c1k/pyenv/lib/python3.11/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context return func(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/k_diffusion/sampling.py", line 852, in sample_dpmpp_3m_sde_gpu return sample_dpmpp_3m_sde(model, x, sigmas, extra_args=extra_args, callback=callback, disable=disable, eta=eta, s_noise=s_noise, noise_sampler=noise_sampler) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/pyt2c1k/pyenv/lib/python3.11/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context return func(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/k_diffusion/sampling.py", line 809, in sample_dpmpp_3m_sde denoised = model(x, sigmas[i] * s_in, **extra_args) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/samplers.py", line 379, in __call__ out = self.inner_model(x, sigma, model_options=model_options, seed=seed) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/samplers.py", line 832, in __call__ return self.predict_noise(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/samplers.py", line 835, in predict_noise return sampling_function(self.inner_model, x, timestep, self.conds.get("negative", None), self.conds.get("positive", None), self.cfg, model_options=model_options, seed=seed) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/samplers.py", line 359, in sampling_function out = calc_cond_batch(model, conds, x, timestep, model_options) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/samplers.py", line 195, in calc_cond_batch return executor.execute(model, conds, x_in, timestep, model_options) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/patcher_extension.py", line 110, in execute return self.original(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/samplers.py", line 308, in _calc_cond_batch output = model.apply_model(input_x, timestep_, **c).chunk(batch_chunks) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/model_base.py", line 130, in apply_model return comfy.patcher_extension.WrapperExecutor.new_class_executor( ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/patcher_extension.py", line 110, in execute return self.original(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/model_base.py", line 159, in _apply_model model_output = self.diffusion_model(xc, t, context=context, control=control, transformer_options=transformer_options, **extra_conds).float() ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/pyt2c1k/pyenv/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl return self._call_impl(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/pyt2c1k/pyenv/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1747, in _call_impl return forward_call(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/ldm/flux/model.py", line 204, in forward out = self.forward_orig(img, img_ids, context, txt_ids, timestep, y, guidance, control, transformer_options, attn_mask=kwargs.get("attention_mask", None)) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/ldm/flux/model.py", line 117, in forward_orig txt = self.txt_in(txt) ^^^^^^^^^^^^^^^^ File "/opt/pyt2c1k/pyenv/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl return self._call_impl(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/pyt2c1k/pyenv/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1747, in _call_impl return forward_call(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/ops.py", line 68, in forward return self.forward_comfy_cast_weights(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/local/.8A/terminal/ComfyAI/ComfyUI/comfy/ops.py", line 64, in forward_comfy_cast_weights return torch.nn.functional.linear(input, weight, bias) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ``` ## System Information - **ComfyUI Version:** unknown - **Arguments:** main.py --highvram --disable-smart-memory --disable-cuda-malloc --listen 127.0.0.1 --auto-launch --verbose ERROR --port 4096 --preview-method latent2rgb --output-directory /opt/local/.8A/terminal/ComfyAI/Downloads/04112025 - **OS:** posix - **Python Version:** 3.11.9 (main, Jul 9 2024, 00:31:01) [GCC 14.1.1 20240522] - **Embedded Python:** false - **PyTorch Version:** 2.5.1+rocm6.3.0 ## Devices - **Name:** cuda:0 AMD Radeon Graphics : native - **Type:** cuda - **VRAM Total:** 25189052416 - **VRAM Free:** 0 - **Torch VRAM Total:** 0 - **Torch VRAM Free:** 0 ```

NeuroCatJan 12, 2025· 5 reactions

CivitAI

This is the best model I have tried. It handles both everyday quick generation tasks and more complex prompts perfectly! And often much better and, most importantly, faster than most 20-step models.

I try other models periodically, but so far I keep coming back to this one.

FluffiesJan 16, 2025

How many steps do you go with?

NeuroCatJan 16, 2025· 1 reaction

@Fluffies It's just four steps.

yndearFeb 9, 2025

CivitAI

It would be cool to look at the NF4 version of the model, is there a similar plan?

princedautomneJul 30, 2025

CivitAI

And the best part, it's Apache 2.0. So thanks a bunch. :D :D :D

Checkpoint