jgrfw/comfyui-aicg3d-skills
Security testingThis page was translated automatically. Technical details may differ from the original. View the original
Independent ComfyUI prompt plugin: MiniMax H3 Skill management, prompt generation, reference material organization, and quality inspection.
Overview
An independent ComfyUI prompt plugin. It adapts the "Comic Drama Lao Li AIGC Creation Full-Process Skill" to the MiniMax H3 version, extending simple creativity into H3-ready prompts with a single master node. - Added "Digital Human Prompt Word Generation" to recognize real lyrics and segmented timecodes from song audio. - Outputs six-segment prompts that can be directly pasted into the 11th MiniMax H3 digital human workflow, including track relationships and fully_copy track relationships and segmented lyrics. - Compatible digital human nodes will delete the actual behavior of tags, avoiding leaving missing fields. - Fixed invalid input errors caused by "Digital Human Prompt Generation" inconsistent with previous quick task values. - Added audio-prompt-reverse digital human-specific skill and added music auditory illusion filtering for lyric recognition. - Added "Quick Tasks": General creation, image reverse deduction prompts, video reverse deduction prompts. - Image reverse inference auto-loading image-prompt-reverse, outputs only one image that can be directly copied and pasted to generate prompts. - Video reverse inference autoloads video-prompt-reverse, outputs only six official fields for MiniMax H3 Ref2VA, without attaching camera timelines or analyses. - Inference mode hides irrelevant parameters and forces actual images/videos and visual models, avoiding offline template fake analysis. - Task mode adds H3 six-parameter standard prompts, text-to-image (T2I), and front-to-end frame generation (FL2I), all using independent output rules. - All Skill final outputs are uniformly cleaned up as prompts that can be directly copied and pasted: image tasks retain only positive prompts, H3 tasks keep only standard fields, and summaries, breakdowns, explanations, and compliance nonsense are deleted. - Added AICG3D SKILLS · Video dialogue recognition, identifying real lines, timecodes, and speaker mapping from original video audio tracks. - Dialogue timeline: Can be directly integrated into H3 prompt inference master node, writing Ref2VA dialogue as S1/S2 -> Picture/Subject. - Character scene replacement skill adds lip movement, language, speech order, and original track reuse rules. - Removes dialogue created by the model during runtime, retaining only verbatim text provided by users or recognition nodes. - Unified the user-visible names, short descriptions, and demo prompts of 9 Skills from Seedance to MiniMax H3. - Retained the original stable directory ID; old workflows, API calls, and serialized data remain compatible. - Synchronously updated T8 source records and Skill license ledgers, clearly marking this interface metadata modification.
Overview (machine translated)
An independent ComfyUI prompt plugin. It adapts the “Comic Drama Lao Li AIGC Creation Full-Process Skill” to the MiniMax H3 version: using a single masternode to extend a line of creativity, reference material, and skill rules into a six-stage prompt that can be directly fed to MiniMax H3.
Core Purpose in One Sentence: You give a creative sentence, a few images/a video/a song, and it generates a six-segment prompt that can be directly copied and pasted to the MiniMax H3.
What is this?
This plugin is responsible for prompt engineering, material organization, skill management, and quality inspection. It does not generate or host video model weights, nor does it grant third-party rights to materials, characters, voices, trademarks, or models.
For the full location description, see docs/PURPOSE.md.
Core Competencies
- Multi-model backend: Supports online APIs (OpenAI-compatible interfaces, Gemini, RunningHub) and local vision models (Transformers / GGUF under
ComfyUI/models/llm). - Quick Tasks: Generate general creation, image reverse-of-the-word prompts, video reverse-deduction prompts, and digital human prompts. After selection, the corresponding skill is automatically loaded and irrelevant parameters are hidden.
- Digital Human Prompt Generation: Recognize real lyrics and timecode segments from song audio, generating prompts that allow the MiniMax H3 digital human to sing word by word along with the original singer, with lip-syncing and lyrics.
- One-sentence multi-material mapping: Enter “Figures 1 and 2 fight scenes in Figure 3” to automatically map
- Built-in MiniMax H3 Director Skills: Automatically loads H3 formats, battles, storyboarding, tone, continuity, and quality control rules.
- Extra 284 Skills from T8 Creative DNA: covering product ads, continuous chases, brand promotions, micro-emoticons, material transformations, character appearances, and other short video mechanisms.
- Skill Pack T8 Case Preview: Pull down to hover to display the poster first, then hover to switch to the corresponding GIF animation.
- Drafts Can Be Generated Without Models: Outputs fully structured offline drafts when no model is configured, making further modifications easier.
- H3 Prompt Check Nodes: Override fields, image numbers, material references, dialogue, sound, and first shot anchoring.
Built-in Skills
| Skill | Purpose |
|---|---|
minimax-h3-short-drama-director |
Default Director Skill, which automatically loads H3 format, storyboarding, tone, continuity, and quality inspection rules. |
image-prompt-reverse |
Image inference: Output a copy of the image to generate prompts. |
This page was translated automatically. Technical details may differ from the original.
Recommended Tools
Try a different keyword or remove a filter.
Install
npx skillfish add jgrfw/comfyui-aicg3d-skills