You just want to get started.
Describe what you need in plain words: text, an image, a voice-over, or a piece of music. One app instead of four tools, and your data never leaves your computer.
04 · KING
AiKing
Local AI, no tinkering required.
Local AI is powerful, but getting there is rough: runners, models, compute libraries, and a different tool for every task. AiKing takes care of that. Your computer is set up in minutes, letting you work with text, image, speech, and music—directly in the app or from your own software.
Screenshot coming soon
- Setup
- in minutes
- Runs
- on your computer
- Tasks
- Text, image, speech, music
- Cost per request
- Zero
What it’s for
Local AI without having to dig into the tech.
Running AI on your own machine means facing endless choices: which software, which model, and will it even fit into memory? AiKing answers these questions for you, sets everything up, and keeps it up to date. What’s left is one central hub for all your tasks.
Who it’s for
You build apps with AI.
AiKing is the interface for your local models, fully OpenAI-compatible. It handles incoming requests, routes them to the right model, and distributes the workload across multiple machines if you have them.
What AiKing can do
Six tasks, one address. You decide which model powers them.
01 · POST /text/generate
Generate text
Use LM Studio, Ollama or oMLX with any model that runs on them.
02 · POST /image/generate
Generate images
AiKing does the processing itself with FLUX.2, right on your computer.
03 · POST /image/recognize
Recognize images
Have images and PDFs described and their contents extracted.
04 · POST /speech/generate
Generate speech
Have text read aloud — the request chooses the voice and speed.
05 · POST /speech/recognize
Recognize speech
Turn recordings into text with Whisper.
06 · POST /music/generate
Generate music
Generate music tracks with or without vocals.
See what it looks like
Screenshot coming soon
Screenshot coming soon
Screenshot coming soon
Get started
- 01
Check your computer
AiKing checks how much memory is available for models and recommends a suitable setup, with alternatives.
- 02
Set up engines
AiKing installs what you need and keeps it up to date.
- 03
Load models
Setup takes a few minutes. Models keep downloading in the background, taking 10 to 60 minutes depending on your connection.
- 04
Get started
Enter your prompt in the app or connect your applications with your own key.
In the app
Four pages, one interface.
Text, Image, Speech, and Music each have their own page: enter a prompt, view the result, and adjust parameters.
- In German – For images and music, AiKing writes the English prompt for you.
- Follow-up questions – With Text, you can continue a conversation, including with PDFs and images attached.
- Versions – Every result is kept and can be compared, even with a different model.
For developers
Existing OpenAI libraries keep working as they are.
Your application specifies a purpose like “text-generate” instead of a model. If you switch models, it won’t notice.
- One key per application – Every request is logged.
- Priority – A queue decides who gets served first.
- In control – Choose your own engine and model.
curl http://localhost:1235/v1/chat/completions \
-H "Authorization: Bearer aik_…" \
-H "Content-Type: application/json" \
-d '{
"model": "text-generate",
"messages": [{ "role": "user",
"content": "Hallo AiKing" }]
}'Multiple computers
Share the workload.
Each computer runs its own AiKing; there’s no boss. Computers on the same network find each other automatically. Otherwise, just enter the address.
- Fixed routing – Requests for a given purpose are sent to another computer.
- Failover – If an engine doesn’t respond or is overloaded, a fallback purpose takes over.
- Load balancing – The request goes wherever the queue is shortest.
Under the hood
Four layers, one control center.
When an AI responds on your computer, four layers work together. AiKing is the top layer: it sets up the layers below it for you and distributes the work. For images, AiKing goes deeper and handles the processing itself with its own AiKing FLUX program.
- AiKing – the control center – Accepts requests from the app or your applications, decides who handles them, and returns the result. Sets up runners and models, keeps them up to date, distributes work across multiple devices, and keeps a log.
- The runner – keeps a model ready – A program that loads a model into memory, accepts requests, and processes them one by one. Examples include LM Studio, Ollama, and oMLX, all free; AiKing can install them for you.
- The compute library – does the calculations – Runs the model on the graphics chip, word by word or pixel by pixel. Examples include MLX and llama.cpp. It’s built into the runner, so you never see it directly.
- The model – what it has learned – A large file containing billions of numbers from training, ranging from a few to several dozen gigabytes. It does nothing on its own. Examples include Qwen, Gemma, FLUX.2, and Whisper. You can swap models at any time.
Frequently asked questions
What buyers often want to know.
- Do I need LM Studio if I have AiKing? – For text, AiKing works with a runner such as LM Studio, Ollama, or oMLX. You don’t have to set it up yourself: AiKing can install it and download a suitable model for you.
- What can AiKing do that LM Studio can’t do on its own? – LM Studio runs text models. AiKing adds image, speech, and music capabilities, a shared interface for all tasks, one key per application, support for multiple devices, and setup.
- Can I keep using models I already have? – Yes. If LM Studio or Ollama is already installed, AiKing finds them and their models and takes them over.
- Does every request cost something? – No. Processing happens on your computer, so there are no per-request fees.
Requirements
- Chip
- Apple chip, M1 or later
- Memory
- From 8 GB for text and speech, 16 GB for images too, 32 GB for music too, 48 GB for the full setup
- Free space
- About 20 to 60 GB, depending on your setup
- Internet
- For setup and downloads
- Network
- Home network or VPN — AiKing shouldn’t be exposed to the internet