Extensions/CodaComfy Nodes
ComfyUI Extension

CodaComfy Nodes

ComfyUI nodes for OpenRouter video, dialogue and music, with cast and frame stacks, a LoRA browser, and a latent chooser.

By codalanguez·Created about 11 hours ago·Updated about 10 hours ago· 0
codalanguez/CodaComfy-Nodes
Nodes—
On cloudLocal install
Stars0
Updatedabout 10 hours ago
Readme

CodaComfy Nodes

A ComfyUI extension for generating video from a prompt, staging on-screen dialogue, and scoring the finished edit. The pack also includes a LoRA browser with stack controls and a latent image chooser.

Install the extension

  1. Download the extension ZIP. Extract the CodaComfy-Nodes folder directly into your ComfyUI installation's custom_nodes directory. The file custom_nodes/CodaComfy-Nodes/__init__.py should exist. If you already run this pack under the name CodaComfyTools, update that installation and keep one copy enabled.
  2. Use Python 3.11 or newer. From the extracted folder, run python -m pip install -r requirements.txt using ComfyUI's Python interpreter. The video workflows use ComfyUI's built-in video, audio, math, and text overlay nodes. Update ComfyUI if one of those nodes is missing.
  3. For the OpenRouter nodes, set OPENROUTER_API_KEY in the environment of the ComfyUI server. Restart ComfyUI and reload the canvas. Paid generation uses the account associated with that key.
  4. Open Templates and look for the Coda pack, or drag a JSON file from example_workflows onto the canvas. To check installation, search the node library for OpenRouter Generate Video. The 01 - Simple Text to Video example needs only a prompt and the server's key.

For a Git installation, run the following from the ComfyUI installation directory, then follow steps 2–4. Managers with an Install via Git URL option can use the same repository URL.

git clone https://github.com/codalanguez/CodaComfy-Nodes.git custom_nodes/CodaComfy-Nodes

Seven example workflows

| Example | What to edit | | --- | --- | | 01 · Simple text to video | Set the prompt, model and shot length for a silent clip. | | 02 · Text to video with generated music | Edit the picture and score prompts. The picture determines the music duration. | | 03 · Text to video with an uploaded song | Upload a song, describe the concept, and choose the shared clip length. The listening model chooses a passage and the edit uses your original recording. | | 04 · Text to video with three speakers | Edit each voice, line and performance cue. Chain more cast cards or disable a speaker. | | 05 · Multiscene text dialogue and music | Direct each scene separately, join them, and adjust the score level under the voices. | | 06 · Multiscene image frames and dialogue | Upload two guides for the first scene and one for the second. Edit the cast or disconnect a frame stack for a text-only shot. | | 07 · Full book trailer | Upload an opening guide and title artwork. Edit three scenes, cast cards, title, byline, spoken title and score. |

Templates 03, 06 and 07 contain placeholder filenames. Choose your own files with the upload controls before running. Download just the seven workflow files.

The Fairest book trailer

This 22-second example of the Full book trailer workflow joins three scenes set in Thornfield with a narrated title card. One score runs beneath the dialogue. Watch the finished trailer.

Download The Fairest workflow and guide image (ZIP). Unzip the bundle, copy thornfield-guide.jpg into your ComfyUI input folder, and open The Fairest - Full Book Trailer.json. Edit the character cards to change who speaks, then change the written and spoken title for your own book.

Working with the nodes

  • Video and models. Use Refresh available models on a generation node to load the provider catalog. Choose a model, then a supported shot length. Model availability, frame support and native audio vary by provider. The credit control shows the balance reported for the configured account.
  • On-screen dialogue. Connect zero or more On-Screen Character cards in speaking order. Each card carries appearance, accent, performance cue and line. Keep $dialogue in the scene text to follow the connected cast automatically. With the cast disconnected, the same node can render a scene without dialogue.
  • Frame guides. Chain Frame Stack cards in story order. One image supplies an opening guide. Consecutive image pairs can generate separate shots and incur separate charges. Choose a model that supports your guide inputs.
  • Sound. Use OpenRouter Generate Sound for a score, OpenRouter Character Dialogue for a separate voice track, and Mix Dialogue and Soundtrack to preserve voices under music. The uploaded-song template sends a compressed listening copy to the selected model.
  • Local tools. LoRA Stack (Preview + Triggers) builds a stack with adjustable strengths. Latent Image Chooser (Pause) pauses a latent batch for a visual selection.

The pack is distributed under the MIT license.