MiniMax H3 I2V with Integrated Prompt Enhancer
This Image-to-Video workflow for MiniMax H3 uses a vision-language model to enhance your prompt before the video generation begins.
Simply load a reference image and write a basic description of what you want to happen. The enhancer analyzes both your image and instructions, then converts them into a detailed prompt structured specifically for MiniMax H3.
It can improve the description of:
Characters and visual elements
Actions and sequence of events
Camera movement and framing
Environment, lighting, and atmosphere
Visual continuity and details that should be preserved
Dialogue in the original language
Ambient sounds, sound effects, and music
The enhanced prompt is automatically sent to MiniMax H3. It is also displayed inside the workflow, allowing you to check exactly what H3 will receive.
In my tests, the resulting videos followed the original instructions much more accurately, especially in scenes involving specific actions, character interactions, camera movements, and dialogue.
The workflow includes a switch to enable or disable the Prompt Enhancer. This allows you to use either the enhanced prompt or your original text without changing any connections.
How to use it
Load your reference image.
Write a simple description of what should happen.
Enable USAR PROMPT ENHANCER?
Run the workflow.
Check the final text in PROMPT FINAL ENVIADO AO H3.
The first run may take longer while the vision-language model is loaded. Generating the enhanced prompt also adds some processing time, but in my tests, the improvement in prompt accuracy and instruction following was absolutely worth it.
The original workflow was preserved, while the enhancer was added as an optional and fully integrated stage.
Description
FAQ
Comments (9)
Haven't looked at the workflow too closely but it seems to be the fastest yet!
yes, its faster, i use an 5060 ti 16gb, so i need it xD
Also, could you please work in some latent upscale in v1.3?
i use another workflow just for upscale, nvidia, do you want so?
Very cool. Any chance of getting a stand alone version, like with just the prompt creating with out video creation?
like a chatgpt? just llm?
@kimimorgam Kinda yea but within Comfyui.
@jefharris confyui is not for this, you will get this very slow hehe
@jefharris confyui is not for this, is it very slow for it
Looks like we don't have an active mirror for this file right now.
CivArchive is a community-maintained index — we catalog mirrors that volunteers upload to HuggingFace, torrents, and other public hosts. Looks like no one has uploaded a copy of this file yet.
Some files do get recovered over time through contributions. If you're looking for this one, feel free to ask in Discord, or help preserve it if you have a copy.