Lyonir Studio

🐺 ComfyUI Lyonir Studio

Custom nodes for ComfyUI, created by Lyonir Studio.

Current public version: v3.5.13.

Included nodes

• 🐺 Lyonir Sampler
• 🐺 Lyonir Resolution
• 🐺 Lyonir Save Image
• 🐺 Lyonir Save Video
• 🐺 Lyonir Image Compare
• 🐺 Lyonir Qwen3-TTS Voice Clone
• 🐺 Lyonir Qwen3-TTS Voice Design
• 🐺 Lyonir Qwen3-TTS Custom Voice

Example workflows

Download a JSON below and drag it into ComfyUI. These examples use the public Lyonir nodes in v3.5.13 or later. Install the pack requirements and ComfyUI-VideoHelperSuite for video output. Models and reference media are not bundled. Screenshots illustrate the author's local workflows; they are not an automated execution guarantee.

Use ComfyUI Manager to identify missing nodes. Depending on the example, your ComfyUI build must provide MiniMax H3, LTX 2.5, ComfyMathExpression and ComfySwitchNode support. Select equivalent model files available in your own installation.

MiniMax H3 first/last-frame video

Download workflow

Model filenames used in this example:

• minimax_h3_video_vae_int8_convrot.safetensors
• minimax_h3_audio_vae_fp32.safetensors
• minimax_h3_fl2va_pruned_int8_convrot.safetensors
• qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors
• Minimax\minimax_h3_fl2v_turbo_8step_v1.0_comfyui_bf16.safetensors

Connect your own first/last-frame images to the conditioning node when using image-guided generation.

MiniMax H3 reference-to-video

Download workflow

Model filenames used in this example:

• minimax_h3_video_vae_int8_convrot.safetensors
• minimax_h3_audio_vae_fp32.safetensors
• minimax_h3_ref2va_pruned_int8_convrot.safetensors
• qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors
• Minimax\minimax_h3_ref2v_turbo_4step_v0.1_comfyui_bf16.safetensors

Replace the two Load Image selections with your own reference images; the named input images are not included.

LTX 2.5 video with latent upscale

Download workflow

Model filenames used in this example:

• ltx-2.5-latent-spatial-upscaler-x2-bf16-1.0.safetensors
• ltx-2.5-22b-distilled-transformer-comfy-int8-convrot.safetensors
• ltx-2.5-video-vae-bf16.safetensors
• ltx-2.5-audio-vae-bf16.safetensors
• gemma4-12b-with-proj-ltx-2.5-comfy-int8-convrot.safetensors

Z-Image Turbo and Lyonir Save Image

Download workflow

Model filenames used in this example:

• z-image-turbo-fp8-aio.safetensors

Running an example

1. Install missing nodes and models, then restart ComfyUI.
2. Drag the downloaded JSON into ComfyUI.
3. Check model selectors, input images, resolution, duration and output filenames.
4. Run the workflow. Model memory requirements depend on your hardware and quantization.

Highlights

Lyonir Save Video

Designed to behave like VideoHelperSuite's Video Combine while adding a persistent per-node video history.

• Delegates final encoding to the installed VHS_VideoCombine.
• Uses the formats and format-specific controls exposed by Video Combine.
• Adds bit-depth handling where the selected format exposes a controllable pixel format.
• Keeps previous generated videos available as selectable thumbnails.
• Uses a persistent per-node history identity so new nodes do not inherit old histories.
• Proportional preview/node resizing.
• Adjustable thumbnail-size slider.
• VideoHelperSuite-compatible hover audio and Sync Preview behavior.
• ProRes presets that manage their own pixel format are delegated to Video Combine instead of aborting the render.

Lyonir Save Image

Persistent image history directly inside the node, with quick access to previously saved images.

Lyonir Sampler

Unified custom-sampling workflow for supported model families. See MULTIMODEL.md for model-family notes.

Qwen3-TTS

Voice Clone, Voice Design and Custom Voice nodes. Review NOTICE and COMMERCIAL_LICENSES.md before commercial deployment.

Requirements

• A working ComfyUI installation.

• Python dependencies listed in requirements.txt.

• ComfyUI-VideoHelperSuite for 🐺 Lyonir Save Video, because that node delegates final video encoding to the installed VHS_VideoCombine.

• Qwen3-TTS voice nodes: install ComfyUI-Qwen-TTS and its dependencies in the same ComfyUI environment. This provides the backend used by the voice nodes. Model files are downloaded or configured separately.

Some nodes may require additional model files or upstream packages depending on the workflow.

Installation

Git

Open a terminal inside ComfyUI/custom_nodes and run:

git clone https://github.com/Lyonir/ComfyUI-Lyonir-Studio.git

Then install the Python requirements using the same Python environment used by ComfyUI:

pip install -r ComfyUI-Lyonir-Studio/requirements.txt

Restart ComfyUI.

Manual installation

1. Download the repository as ZIP.
2. Extract it.
3. Make sure the final folder is ComfyUI/custom_nodes/ComfyUI-Lyonir-Studio/.
4. Install requirements.txt using the same Python environment as ComfyUI.
5. Restart ComfyUI.

Updating

If installed with Git:

cd ComfyUI/custom_nodes/ComfyUI-Lyonir-Studio
git pull

Restart ComfyUI. If frontend files changed, refresh the browser as well.

Troubleshooting

If a node does not appear after installation:

1. Check the ComfyUI terminal for Python import errors.
2. Confirm the repository folder is directly under ComfyUI/custom_nodes/.
3. Install requirements.txt using the same Python environment as ComfyUI.
4. Restart ComfyUI completely.
5. Refresh the browser frontend.

For Lyonir Save Video, also confirm that ComfyUI-VideoHelperSuite is installed and VHS_VideoCombine is available.

License

The Lyonir Studio code in this repository is distributed under the license included in LICENSE.

Model weights, third-party packages, voice data and upstream projects may have their own licenses and usage restrictions. See NOTICE and COMMERCIAL_LICENSES.md.

Creator

Lyonir Studio

• Website: https://www.lyonirstudio.com/
• Portfolio: https://lyonirstudios.myportfolio.com/