Instructions to use vidfom/Ltx-3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- llama-cpp-python
How to use vidfom/Ltx-3 with llama-cpp-python:
# !pip install llama-cpp-python from llama_cpp import Llama llm = Llama.from_pretrained( repo_id="vidfom/Ltx-3", filename="ComfyUI/models/text_encoders/gemma-3-12b-it-qat-UD-Q4_K_XL.gguf", )
llm.create_chat_completion( messages = "No input example has been defined for this model task." )
- Notebooks
- Google Colab
- Kaggle
- Local Apps
- llama.cpp
How to use vidfom/Ltx-3 with llama.cpp:
Install from brew
brew install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama-server -hf vidfom/Ltx-3:UD-Q4_K_XL # Run inference directly in the terminal: llama-cli -hf vidfom/Ltx-3:UD-Q4_K_XL
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama-server -hf vidfom/Ltx-3:UD-Q4_K_XL # Run inference directly in the terminal: llama-cli -hf vidfom/Ltx-3:UD-Q4_K_XL
Use pre-built binary
# Download pre-built binary from: # https://github.com/ggerganov/llama.cpp/releases # Start a local OpenAI-compatible server with a web UI: ./llama-server -hf vidfom/Ltx-3:UD-Q4_K_XL # Run inference directly in the terminal: ./llama-cli -hf vidfom/Ltx-3:UD-Q4_K_XL
Build from source code
git clone https://github.com/ggerganov/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-server llama-cli # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama-server -hf vidfom/Ltx-3:UD-Q4_K_XL # Run inference directly in the terminal: ./build/bin/llama-cli -hf vidfom/Ltx-3:UD-Q4_K_XL
Use Docker
docker model run hf.co/vidfom/Ltx-3:UD-Q4_K_XL
- LM Studio
- Jan
- Ollama
How to use vidfom/Ltx-3 with Ollama:
ollama run hf.co/vidfom/Ltx-3:UD-Q4_K_XL
- Unsloth Studio new
How to use vidfom/Ltx-3 with Unsloth Studio:
Install Unsloth Studio (macOS, Linux, WSL)
curl -fsSL https://unsloth.ai/install.sh | sh # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for vidfom/Ltx-3 to start chatting
Install Unsloth Studio (Windows)
irm https://unsloth.ai/install.ps1 | iex # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for vidfom/Ltx-3 to start chatting
Using HuggingFace Spaces for Unsloth
# No setup required # Open https://huggingface.co/spaces/unsloth/studio in your browser # Search for vidfom/Ltx-3 to start chatting
- Docker Model Runner
How to use vidfom/Ltx-3 with Docker Model Runner:
docker model run hf.co/vidfom/Ltx-3:UD-Q4_K_XL
- Lemonade
How to use vidfom/Ltx-3 with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull vidfom/Ltx-3:UD-Q4_K_XL
Run and chat with the model
lemonade run user.Ltx-3-UD-Q4_K_XL
List all available models
lemonade list
File size: 6,625 Bytes
e00eceb | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 | from typing_extensions import override
from comfy_api.latest import IO, ComfyExtension, Input
from comfy_api_nodes.apis.wavespeed import (
FlashVSRRequest,
TaskCreatedResponse,
TaskResultResponse,
SeedVR2ImageRequest,
)
from comfy_api_nodes.util import (
ApiEndpoint,
download_url_to_video_output,
poll_op,
sync_op,
upload_video_to_comfyapi,
validate_container_format_is_mp4,
validate_video_duration,
upload_images_to_comfyapi,
get_number_of_images,
download_url_to_image_tensor,
)
class WavespeedFlashVSRNode(IO.ComfyNode):
@classmethod
def define_schema(cls):
return IO.Schema(
node_id="WavespeedFlashVSRNode",
display_name="FlashVSR Video Upscale",
category="api node/video/WaveSpeed",
description="Fast, high-quality video upscaler that "
"boosts resolution and restores clarity for low-resolution or blurry footage.",
inputs=[
IO.Video.Input("video"),
IO.Combo.Input("target_resolution", options=["720p", "1080p", "2K", "4K"]),
],
outputs=[
IO.Video.Output(),
],
hidden=[
IO.Hidden.auth_token_comfy_org,
IO.Hidden.api_key_comfy_org,
IO.Hidden.unique_id,
],
is_api_node=True,
price_badge=IO.PriceBadge(
depends_on=IO.PriceBadgeDepends(widgets=["target_resolution"]),
expr="""
(
$price_for_1sec := {"720p": 0.012, "1080p": 0.018, "2k": 0.024, "4k": 0.032};
{
"type":"usd",
"usd": $lookup($price_for_1sec, widgets.target_resolution),
"format":{"suffix": "/second", "approximate": true}
}
)
""",
),
)
@classmethod
async def execute(
cls,
video: Input.Video,
target_resolution: str,
) -> IO.NodeOutput:
validate_container_format_is_mp4(video)
validate_video_duration(video, min_duration=5, max_duration=60 * 10)
initial_res = await sync_op(
cls,
ApiEndpoint(path="/proxy/wavespeed/api/v3/wavespeed-ai/flashvsr", method="POST"),
response_model=TaskCreatedResponse,
data=FlashVSRRequest(
target_resolution=target_resolution.lower(),
video=await upload_video_to_comfyapi(cls, video),
duration=video.get_duration(),
),
)
if initial_res.code != 200:
raise ValueError(f"Task creation fails with code={initial_res.code} and message={initial_res.message}")
final_response = await poll_op(
cls,
ApiEndpoint(path=f"/proxy/wavespeed/api/v3/predictions/{initial_res.data.id}/result"),
response_model=TaskResultResponse,
status_extractor=lambda x: "failed" if x.data is None else x.data.status,
poll_interval=10.0,
max_poll_attempts=480,
)
if final_response.code != 200:
raise ValueError(
f"Task processing failed with code={final_response.code} and message={final_response.message}"
)
return IO.NodeOutput(await download_url_to_video_output(final_response.data.outputs[0]))
class WavespeedImageUpscaleNode(IO.ComfyNode):
@classmethod
def define_schema(cls):
return IO.Schema(
node_id="WavespeedImageUpscaleNode",
display_name="WaveSpeed Image Upscale",
category="api node/image/WaveSpeed",
description="Boost image resolution and quality, upscaling photos to 4K or 8K for sharp, detailed results.",
inputs=[
IO.Combo.Input("model", options=["SeedVR2", "Ultimate"]),
IO.Image.Input("image"),
IO.Combo.Input("target_resolution", options=["2K", "4K", "8K"]),
],
outputs=[
IO.Image.Output(),
],
hidden=[
IO.Hidden.auth_token_comfy_org,
IO.Hidden.api_key_comfy_org,
IO.Hidden.unique_id,
],
is_api_node=True,
price_badge=IO.PriceBadge(
depends_on=IO.PriceBadgeDepends(widgets=["model"]),
expr="""
(
$prices := {"seedvr2": 0.01, "ultimate": 0.06};
{"type":"usd", "usd": $lookup($prices, widgets.model)}
)
""",
),
)
@classmethod
async def execute(
cls,
model: str,
image: Input.Image,
target_resolution: str,
) -> IO.NodeOutput:
if get_number_of_images(image) != 1:
raise ValueError("Exactly one input image is required.")
if model == "SeedVR2":
model_path = "seedvr2/image"
else:
model_path = "ultimate-image-upscaler"
initial_res = await sync_op(
cls,
ApiEndpoint(path=f"/proxy/wavespeed/api/v3/wavespeed-ai/{model_path}", method="POST"),
response_model=TaskCreatedResponse,
data=SeedVR2ImageRequest(
target_resolution=target_resolution.lower(),
image=(await upload_images_to_comfyapi(cls, image, max_images=1))[0],
),
)
if initial_res.code != 200:
raise ValueError(f"Task creation fails with code={initial_res.code} and message={initial_res.message}")
final_response = await poll_op(
cls,
ApiEndpoint(path=f"/proxy/wavespeed/api/v3/predictions/{initial_res.data.id}/result"),
response_model=TaskResultResponse,
status_extractor=lambda x: "failed" if x.data is None else x.data.status,
poll_interval=10.0,
max_poll_attempts=480,
)
if final_response.code != 200:
raise ValueError(
f"Task processing failed with code={final_response.code} and message={final_response.message}"
)
return IO.NodeOutput(await download_url_to_image_tensor(final_response.data.outputs[0]))
class WavespeedExtension(ComfyExtension):
@override
async def get_node_list(self) -> list[type[IO.ComfyNode]]:
return [
WavespeedFlashVSRNode,
WavespeedImageUpscaleNode,
]
async def comfy_entrypoint() -> WavespeedExtension:
return WavespeedExtension()
|