Self-Hosted AI for WordPress — Your Keys, Your Hardware, Your Site
Hosted AI is the fastest way to start, but it isn’t the only way to run. StudioUI Pro turns the same WordPress studio into a fully self-hosted AI stack: bring your own API keys from the providers you already trust, or point the plugin at ComfyUI and LM Studio running on your own hardware. No per-generation fees, no usage meters, and prompts that never have to leave infrastructure you control.
Two Ways to Own Your AI Stack
“Self-hosted AI” means different things to different people, and StudioUI Pro supports both readings:
1. Bring your own API keys (BYOK)
Keep using cloud models, but on your accounts. Pro lets you plug in your own keys from OpenAI, FLUX/BFL, fal.ai, Groq, Mistral, DeepSeek and more. You pay each provider directly at their rates, StudioUI adds no per-generation fee on top, and a multi-key pool with automatic failover keeps generation running if one key hits a rate limit. If you’ve searched for a bring your own API key AI image generator, this is exactly that — wrapped in a full WordPress studio.
2. Self-host the backends themselves
Go further and run the models on hardware you own. StudioUI connects to ComfyUI for image and media generation, and to LM Studio, Ollama or any OpenAI-compatible endpoint for chat. Your GPU box, home server or rented cloud GPU does the work; WordPress is the clean front end your users see. Prompts and outputs stay inside your own network.
Why Self-Host AI on WordPress at All?
- No per-generation fees. On your own keys or hardware there is no usage meter. Generate ten images or ten thousand — the licence cost is the same flat $149/yr.
- Privacy and control. With self-hosted ComfyUI and LM Studio backends, prompts and generated media never have to leave your infrastructure. For client work, internal tools or regulated niches, that’s often the whole requirement.
- Model freedom. ComfyUI opens the full open-weights ecosystem — run the checkpoints, LoRAs and workflows you choose, not a fixed menu. Pro also unlocks StudioUI’s full model catalog on the API side.
- Predictable cost at volume. Pay-as-you-go credits are ideal for bursty use; heavy daily volume is usually cheaper on your own GPU or negotiated provider rates. Pro lets you switch when the math says so.
- No lock-in. Keys, hardware and WordPress site are all yours. StudioUI is the studio layer, not a platform holding your workflow hostage.
What the Self-Hosted Stack Looks Like
A typical Pro setup has three layers, all of which you control:
WordPress + StudioUI (the front end)
The plugin renders the studio — image, video, GIF, music and chat companion generation — inside your site, with white-label branding by default. Members log in to your domain and generate; results land in your Media Library.
ComfyUI (images and media)
ComfyUI is the workhorse of self-hosted image generation: node-based workflows, support for FLUX-class and SDXL-class open models, LoRAs and custom pipelines. StudioUI speaks to it directly — see how to connect ComfyUI to WordPress or start with what ComfyUI is if it’s new to you. A step-by-step is in how to self-host Stable Diffusion for WordPress.
LM Studio / Ollama (chat and companions)
For chat, StudioUI talks to any OpenAI-compatible local runtime. LM Studio and Ollama both run modern open-weight LLMs on consumer hardware and expose a local API endpoint — paste it into StudioUI’s settings and your chat companions run entirely on your machine.
How to Set It Up
1. Install StudioUI free
Download the plugin and activate it on any self-hosted WordPress site. You can start on the Free tier’s hosted credits today and move to self-hosting whenever you’re ready — the studio your users see doesn’t change.
2. Go Pro
Pro is $149/yr and unlocks BYO keys, self-hosted backends, the full model catalog, higher generation limits and multi-key failover. See pricing for the full comparison.
3. Connect your keys or backends
In the plugin settings, add provider API keys (see using your own API key in WordPress), or point StudioUI at your ComfyUI and LM Studio/Ollama endpoints. Mix and match: many sites run local image generation on a GPU box and route chat to a cheap fast API, or the reverse.
Three Common Setups
- All-cloud BYOK: no hardware at all — provider keys for image, video and chat, paid at provider rates, with StudioUI’s multi-key failover smoothing rate limits. The right shape for agencies standardising across many client sites.
- Hybrid: a local ComfyUI box handles image generation (the highest-volume, most price-sensitive modality) while chat rides a fast cheap API. Most self-hosters end up here.
- Fully local: ComfyUI for media plus LM Studio or Ollama for chat, everything on your own metal. Maximum privacy, zero marginal cost per generation.
Hosted vs Self-Hosted, Honestly
| Free (hosted credits) | Pro (BYOK / self-host) | |
|---|---|---|
| Setup | One-time connect code, ~1 minute | Your keys or your ComfyUI/LM Studio endpoints |
| Cost shape | Pay per generation in credits ($1 = 1,000 credits) | $149/yr flat + whatever you pay providers or your own hardware |
| Hardware needed | None | None for BYOK; a GPU for local ComfyUI |
| Model choice | Curated hosted lineup | Full catalog + any open-weights model you run |
| Data path | Via StudioUI’s hosted gateway | Direct to your providers, or never leaves your network |
| Best for | Getting started, bursty usage | Volume, privacy, model control, client work |
There’s no forced migration in either direction — Free and Pro use the same plugin and the same studio, so upgrading is a settings change, not a rebuild. If you stop renewing Pro, the site simply reverts to the hosted model.
Who Goes Self-Hosted
- Developers and agencies who manage provider relationships themselves and put a white-label AI studio on client sites with no per-generation cost exposure.
- Owners of GPU hardware who already run ComfyUI or a local LLM and want a polished, member-ready front end instead of sharing a raw web UI.
- Privacy-sensitive projects where prompts and outputs must stay on infrastructure you control.
- High-volume sites where flat-cost generation beats any pay-per-use meter. If you’re reselling generation to members, self-hosting also maximises the margin — see selling AI credits on your website.
FAQ
Do I need a GPU to use StudioUI Pro?
No. BYO API keys is fully cloud-based — you just pay providers directly instead of using credits. A GPU only comes into play if you choose to run ComfyUI or a local LLM yourself.
Which providers can I bring keys from?
OpenAI, FLUX/BFL, fal.ai, Groq, Mistral, DeepSeek and more — plus any OpenAI-compatible endpoint for chat, which covers LM Studio, Ollama and most self-hosted inference servers.
Can I mix hosted credits and my own keys?
Yes. Pro includes everything in Free, so a site can run some modalities on your own backends and fall back to hosted credits elsewhere — useful while you migrate or when your GPU is busy.
Is self-hosted generation private?
When you run ComfyUI and LM Studio/Ollama locally, generation happens on your hardware and StudioUI doesn’t need to route your prompts through its gateway. With BYO cloud keys, data goes directly to the provider you chose under their terms.
What happens if I cancel Pro?
Your licence runs to the end of the paid year, then the site reverts automatically to the Free hosted model. Nothing breaks and your studio stays up.
Run AI on your own terms
Install StudioUI free today, then go Pro when you’re ready to plug in your keys or your hardware — $149/yr, no per-generation fees, white-label by default.