Save 4 hours! We auto-setup your workflow! Free!

Drop your workflow.json — we handle every dependency, custom node, and model. Just open the link and run.

Auto-Setup Workflow Json (Free) Now!
ComfyUI > Nodes > ComfyUI-Prompt-Rewriters > H3 Ref2V Enhancer

ComfyUI Node: H3 Ref2V Enhancer

Class Name

ConnorH3R2VEnhancer

Category
Prompt_Enhancement
Author
connorpietrasik (Account age: 2593days)
Extension
ComfyUI-Prompt-Rewriters
Latest Updated
2026-08-19
Github Stars
0.0K

How to Install ComfyUI-Prompt-Rewriters

Install this extension via the ComfyUI Manager by searching for ComfyUI-Prompt-Rewriters
  • 1. Click the Manager button in the main menu
  • 2. Select Custom Nodes Manager button
  • 3. Enter ComfyUI-Prompt-Rewriters in the search bar
After installation, click the Restart button to restart ComfyUI. Then, manually refresh your browser to clear the cache and access the updated list of nodes.

Visit ComfyUI Online for ready-to-use ComfyUI environment

  • Free trial available
  • 16GB VRAM to 80GB VRAM GPU machines
  • 400+ preloaded models/nodes
  • Freedom to upload custom models/nodes
  • 200+ ready-to-run workflows
  • 100% private workspace with up to 200GB storage
  • Dedicated Support

Run ComfyUI Online

H3 Ref2V Enhancer Description

Rewrites video ideas and reference images into structured Minimax H3 full-reference prompts.

H3 Ref2V Enhancer:

ConnorH3R2VEnhancer rewrites a video idea into a structured prompt intended for Minimax H3 full-reference video generation. It sends your idea and any supplied reference images to a chat-completions service, along with detailed instructions for describing subjects, reference relationships, shots, camera movement, dialogue, sound, and music. This can help you turn a brief concept into a more explicit production-style prompt. The node does not generate or validate a video; the quality and accuracy of the rewrite depend on the service you connect and the information you provide.

H3 Ref2V Enhancer Input Parameters:

initial_prompt

Required STRING containing the video idea you want rewritten. The field supports multiple lines, so you can describe the intended action, duration, visual style, dialogue, sound, and how any references should be used. Those details give the connected service a clearer basis for its rewrite; details you leave out may not be represented as you intended. The default is (3 second video) A magic cat shooting lightning at skeletons. No minimum or maximum length is specified in the node source.

llm_url

Optional STRING specifying the chat-completions endpoint that receives the rewrite request. The default is http://localhost:8080/v1/chat/completions. Use an endpoint reachable from the machine running the node and compatible with the request and response format described below; changing this address changes which service performs the rewrite. The field is single-line, and the source specifies no minimum or maximum length or list of allowed URLs. The node sends a JSON request containing messages and expects a response with choices[0].message.content. It does not add an authorization header.

ref_images

Optional IMAGE input for visual references to include with your prompt. When connected, the node converts each image it receives into a PNG and sends it to the chat-completions endpoint as an inline image alongside your text. This can help the service describe referenced visual content, provided that endpoint accepts image content. The images serve as references for rewriting; this node does not use them to generate video. There is no default image, and the source specifies no minimum or maximum image count. The conversion expects image data that can be saved as RGB.

extra_start_note

Optional single-line STRING added near the beginning of the instructions sent to the service. Use it for a concise additional requirement that should guide the rewrite, such as a constraint on how to describe the scene. It is inserted directly before the node’s built-in guide, so clear wording and a trailing space or punctuation can make the combined instruction easier to read. When omitted, no note is added. No default text, minimum length, or maximum length is specified.

extra_end_note

Optional single-line STRING added as a final bullet in the instructions sent to the service. Use it for a specific finishing constraint, such as a reminder to preserve a detail from your concept. When omitted, no final bullet is added. No default text, minimum length, or maximum length is specified.

H3 Ref2V Enhancer Output Parameters:

improved_prompt

STRING containing the text returned by the connected chat-completions service as its first choice’s message content. The node’s instructions ask that service to produce a full-reference rewrite with subject_definitions, summary, retention_analysis, detailed_description, overall_soundscape, and non_diegetic_music sections. Treat this output as a proposed prompt to review before video generation: the node returns the service’s text without checking that every requested section is present or that its descriptions accurately match your references.

H3 Ref2V Enhancer Usage Tips:

  • Start with the action and intended duration, then add any essential visual, dialogue, and sound details. If exact spoken words or visible text matter, include them explicitly so the rewrite has the wording to preserve.
  • Connect ref_images when appearance or composition matters, and explain in initial_prompt what each image is meant to guide. An image reference alone does not tell the service whether you intend it as a character reference, scene reference, or target frame.
  • Confirm that llm_url points to a running service that accepts chat messages with inline image_url content before using ref_images. The default address refers to a service on the node’s local machine; the node does not start that service.
  • Use extra_start_note or extra_end_note for short, task-specific constraints. Review improved_prompt afterward, particularly its reference labels and any claims about audio or video assets, because this node supplies images and text but does not itself attach audio or video files.

H3 Ref2V Enhancer Common Errors and Solutions:

ConnectionRefusedError or a connection error mentioning localhost:8080

  • Explanation: The node could not connect to the endpoint in llm_url. With the default address, a common cause is that no service is listening on port 8080.
  • Solution: Start your compatible chat-completions service, confirm its address and port, and update llm_url if it runs elsewhere.

A timeout error after waiting for the response

  • Explanation: The request did not complete within the node’s 600-second timeout. A busy service or large reference images can contribute to a slow request.
  • Solution: Check that the service is responding, try a request without images to isolate the issue, and reduce image size or service load where practical.

JSONDecodeError while reading the response

  • Explanation: The endpoint returned content that could not be parsed as JSON, such as an HTML error page or an empty response. The node attempts to parse the response without first checking its HTTP status.
  • Solution: Verify the endpoint path and inspect the service’s response or logs. It must return JSON in the expected chat-completions format.

KeyError: 'choices', KeyError: 'message', or KeyError: 'content'

  • Explanation: The response was JSON, but it did not contain the fields the node reads at choices[0].message.content. This can also happen when the service returns a JSON error instead of a completion.
  • Solution: Check the endpoint’s response and service logs. Use a compatible endpoint and resolve any reported request errors.

IndexError: list index out of range

  • Explanation: The response contained an empty choices list, but the node tries to read its first item.
  • Solution: Check why the service returned no completion and confirm that it produces at least one choice for this request.

An image conversion error mentioning RGB or array shape

  • Explanation: The node converts each reference image to an RGB PNG. An image with an incompatible shape or channel layout may fail during that conversion.
  • Solution: Supply a standard RGB image through ref_images and test again with one image before adding more.

The service rejects the request or returns an authentication error

  • Explanation: The node sends JSON containing messages but does not supply an authorization header. Some endpoints also reject inline image content or require a different request body.
  • Solution: Use an endpoint compatible with this request format and its authentication setup. If your service requires fields or headers the node does not send, the node source would need to be changed.

The rewrite is incomplete or does not follow the requested format

  • Explanation: The node instructs the service to use a six-section format, but it does not validate or repair the returned text.
  • Solution: Review improved_prompt before using it. Make initial_prompt more specific, simplify conflicting notes, and check whether the connected service can reliably follow the supplied instructions.

H3 Ref2V Enhancer Related Nodes

Go back to the extension to check out more related nodes.
ComfyUI-Prompt-Rewriters
RunComfy
Copyright 2025 RunComfy. All Rights Reserved.

RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Models, enabling artists to harness the latest AI tools to create incredible art.

H3 Ref2V Enhancer