🌈RH全能图片@炮老师的小课堂

RunningHub (runninghub.cn) is a Chinese cloud platform that rents GPU-backed AI apps and exposes them through an API - one of the bigger names in that corner of the world, with a real Reddit footprint (~350 corpus mentions) as an affordable ComfyUI-hosting/API alternative. DapaoRHAllImageNode is the single-node wrapper for RunningHub's "全能图片" (all-capable image) series: text-to-image, image-to-image, and multi-reference generation through RunningHub's API, without you renting a single GPU hour yourself.

The inputs that matter

• 🌐 API渠道 - 国内版 (domestic/China) vs 国外版 (international, runninghub.ai). The tooltip is blunt: they use different API addresses and different keys, and the keys are not interchangeable. Pick the one matching the account you registered.
• 🔑 API密钥 - your RunningHub API key. Tooltip again: used for this request only, never written to a file.
• 🤖 模型 - three entries: 全能图片G-2 (default), V2, PRO. The endpoint is chosen by channel + mode combo, per the tooltip.
• 🏷️ 渠道 - 官方稳定版 (official-stable, quality-first) vs 低价渠道版 (low-price channel, default here). The naming is the tradeoff: the stable channel costs more and drifts less; the cheap channel is, well, cheap.
• 🔀 模式 - 文生图 (text-to-image) vs 图生图 (image-to-image). Tooltip: img2img requires at least one reference image wired in.
• 📝 提示词 - your prompt. 📐 画面比例 (21 presets incl. 模型默认), 🧩 分辨率 (模型默认/1k/2k/4k), 🎨 画质 (模型默认/low/medium/high - with an honest tooltip that quality only takes effect on the G-2 official-stable endpoint; other endpoints silently ignore it).
• 🎲 随机种 - cache-only, never sent to RunningHub.
• Optional: 🖼️ 图像1–图像10 reference images, 📋 额外参数JSON escape hatch, and polling controls - 🔁 最大轮询秒数 (1200), ⏱️ 轮询间隔 (5s), ⌛ 请求超时 (60s).

How it works

This is an async-task wrapper: submit to RunningHub's openapi/v2 endpoint, poll the task ID until it lands, fetch the image, decode to an IMAGE tensor. Outputs: 🖼️ 图像 (the tensor for your graph), 🔗 图片链接 (the URL), 📋 响应信息 (raw response for debugging).

The honest read

RunningHub is a legitimate platform with a niche but real international presence, and this node is a competent client for it. What it is not is a way to dodge the genre's costs: every run is a metered API call billed against your RunningHub balance (the README is an advertisement for its membership/invite-code model - the author gets a referral bounty, so treat the "绑定送1000算力" as marketing). Your prompts and images leave your machine to RunningHub's servers, same as every API wrapper (external-api-nodes.md covers the category caution).

Gotchas

• 国内版 vs 国外版 key mismatch is the #1 support question waiting to happen. Your runninghub.cn key won't work on runninghub.ai and vice versa.
• Quality setting is a no-op on the low-price channel and most non-G-2 endpoints. Don't chase 画质=high and wonder why nothing changed.
• Img2img needs an image. Wire a reference or the request will fail - the mode doesn't auto-switch like the dapaoAI image nodes do.
• Polling ≠ instant. 720p video-adjacent jobs can take minutes; the 1200s cap is generous, the 60s request timeout is only for the HTTP call, not the job.
• Chinese-only labels, as with the whole pack.

Install: ComfyUI Manager → "dapaoAPI", or git clone https://github.com/paolaoshi/ComfyUI-dapaoAPI.git into ComfyUI/custom_nodes/, pip install -r requirements.txt, restart. Get a RunningHub account and API key first - the node is useless without one.