T8mars/Comfyui-zhenzhen

What is ComfyUI‑zhenzhen?

ComfyUI‑zhenzhen is a collection of ready‑made ComfyUI nodes that let you call a large number of commercial AI generation APIs (image, video, audio, 3‑D, etc.) directly from the visual workflow engine ComfyUI. The repository bundles:

• Node definitions for dozens of models – e.g. Gemini‑3.1, Flux, Minimax‑H3, Seedream, Wan, Omni, GK‑V2, FlowMusic, FlashVSR, etc.
• Pre‑configured “low‑price” nodes that automatically point to the author’s “Zhenzhen AI Shop” (a non‑profit API marketplace with discounted rates for Chinese and overseas users).
• Example workflows for each node, showing how to generate images, edit them, create videos, upscale, do audio‑to‑video, 3‑D reconstruction, segmentation, region editing, music generation, etc.
• Utility scripts for concurrent submission, robust media download, seed handling, automatic domain fallback, and OSS (object‑storage) configuration.

In short, it turns the ComfyUI visual programming interface into a front‑end for a large catalog of paid AI services, while handling API keys, rate‑limiting, and result downloading for you.

---

Who might use it?

User type	Why it helps them
ComfyUI hobbyists / artists	No need to write custom Python wrappers – just drop a node into the canvas and start generating with the author’s cheap API endpoints.
Chinese developers	Nodes default to the “国内版平价AI小铺” which offers lower‑priced domestic models and a built‑in button to obtain a local API key.
International users	Nodes can switch to the overseas endpoint (ai.t8star.org → ai.t8star.cn fallback) and support many foreign models (e.g., Gemini, Flux, Midjourney).
Workflow builders	Hundreds of example workflows (image‑to‑image, video‑to‑video, audio‑driven avatars, 3‑D generation) are provided, ready to be customized.
Researchers testing many models	The same UI can call dozens of different providers without changing code, making rapid comparison easy.

---

How does it work?

1. API Settings node – stores the base URL, API key, and optional OSS bucket. The UI shows a button that opens the appropriate “Get API key” page for the chosen region.
2. Model‑specific node – each node bundles the model ID, required inputs (text prompt, reference images, video URLs, audio files, etc.) and exposes the same set of controls (seed, size, aspect ratio, NSFW check, etc.). Switching the model inside the node automatically updates the visible controls.
3. Submission & polling – when the node runs, it uploads any media, sends a JSON request to the remote service, then polls for completion. A robust downloader retries transient errors, respects a minimum 120 s timeout, and can fall back to a different CDN domain.
4. Concurrent execution – the repo provides “Concurrent” nodes that split a batch of up to 30 images or 10 videos across parallel API calls, dramatically speeding up large jobs.
5. Result handling – downloaded media are converted into native ComfyUI data types (IMAGE, AUDIO, VIDEO, FILE_3D_GLB). Nodes also output the raw API response (task ID, full JSON) for debugging.

---

What’s new (selected recent updates)

Date	Highlights
2026‑10‑12	Added low‑price image models (zhenzhen-image‑nb‑pro‑lowprice, zhenzhen-image‑nb‑2‑lowprice) and voice‑clone model.
2026‑09‑09	Introduced a 6‑in‑1 GPT‑image node supporting 6 resolution/quality combos; added video‑upscale node FlashVSR.
2026‑08‑30	New Omni‑1.1 Flash video node supporting text‑to‑video, single‑frame, 1/3‑image, and reference‑video modes.
2026‑08‑22	Added 3‑D generation nodes (hunyuan3d‑v3.1‑text‑to‑3d, ‑image‑to‑3d) with GLB output.
2026‑08‑21	FlowMusic node (9‑in‑1) for music generation, lyric writing, audio upload, remix, stems export, etc.
2026‑08‑18	Video super‑resolution node zhenzhen‑FlashVSR‑video‑upscale‑lowprice.
2026‑08‑14	Minimax‑H3 audio‑driven video nodes (fast mode) added.
2026‑07‑31	Minimax H3 “海螺” video node introduced.
2026‑07‑25	Midjourney‑style node covering 16 operations (imagine, blend, upscale, variation, video, etc.).
2026‑07‑10	First release of the “国内版平价AI小铺” API with discounted pricing.

---

How to get started

1. Install ComfyUI (follow the official guide).
2. Clone this repo into the custom_nodes folder of ComfyUI.
3. Open ComfyUI, you’ll see a new category “Zhenzhen” with dozens of nodes.
4. Add an API Settings node, click the “获取国内版 APIKEY” (or overseas) button, paste the key.
5. Drag any model node (e.g., zhenzhen‑image‑g‑v2‑lowprice) onto the canvas, connect a Prompt string, set size/aspect, and run.
6. Use the provided example workflow files (found in the workflow directory) to see full pipelines – e.g., generate → segment → region‑edit → save.

---

Why it matters

• Cost‑effective access – The author negotiates bulk pricing with model providers and passes the savings to users (≈ 7‑8 % discount vs. overseas rates).
• Unified UI – No need to learn each provider’s SDK; everything lives inside the visual ComfyUI graph.
• Rapid prototyping – Switch models, change modalities (image ↔ video ↔ audio) with a single click, making experimentation fast.
• Community‑ready – All nodes are open‑source, and the repo includes detailed changelogs, troubleshooting tips, and video tutorials.

---

TL;DR

ComfyUI‑zhenzhen is an open‑source extension that adds dozens of ready‑to‑use ComfyUI nodes for a wide range of commercial AI generation APIs (image, video, audio, 3‑D, music, etc.). It handles API keys, submission, polling, concurrent execution, and robust downloading, letting artists and developers build complex multimodal workflows without writing any code. The project is actively maintained with frequent model additions and detailed example workflows.