Frequently Asked Questions
Everything you need to know about our inference platform, pay-as-you-go billing, commercial usage, and technical integration.
What is LumozAI?
LumozAI is a serverless AI inference platform where developers and creators can test, run, and scale 70+ frontier open-source and proprietary models with zero infrastructure management. We offer unified API endpoints, web playgrounds, and pay-as-you-go GPU compute.
Who can use LumozAI?
LumozAI is engineered for:
- Indie Hackers & Founders: Ship generative features into your SaaS in minutes.
- Digital Agencies & Studios: High-throughput media generation for marketing campaigns.
- AI Engineers & Researchers: Benchmark and evaluate frontier weights without local hardware limits.
How do I get started?
- Create a free account or sign in.
- Browse the Model Library or test in the AI Studio.
- Retrieve your API key from the Dashboard.
- Send your first inference request in cURL or Python using our API Docs.
Is there a mobile application?
LumozAI is fully responsive and optimized for mobile and tablet browsers, allowing you to run models and review outputs on any device. Native iOS and Android developer companions are on our roadmap.
What types of models are available?
We host a comprehensive roster of multimodal frontier architectures:
- Image Generation: Flux.1 (Dev/Schnell), Stable Diffusion 3.5, HiDream, Midjourney styles.
- Video Generation: CogVideoX, Animatediff, image-to-video synthesis pipelines.
- Audio & Speech: Text-to-speech, voice cloning, and audio enhancement.
- Image-to-Image & Restoration: Background removal, 4x upscaling, face restoration, and clothes swap.
Can I use generated assets for commercial purposes?
Yes. All pay-as-you-go API calls and web runs include full commercial usage rights for generated media, subject to individual base model licenses (e.g. Apache 2.0, OpenRAIL-M).
How are models benchmarked for quality?
Every model undergoes automated cold-start latency analysis, memory footprint profiling, and prompt adherence testing before being promoted to our production API.
Can I request a custom open-source model?
Yes! We routinely add trending open-source architectures from Hugging Face and GitHub. Send your request via our contact page.
How does model deployment work on LumozAI?
You can deploy LoRA weights, fine-tunes, or custom checkpoints directly by providing the Hugging Face repo ID or direct safetensors URL. Our engine downloads the weights and provisions serverless inference endpoints automatically.
Can I keep my deployed models private?
Yes. You can toggle your deployments between Private (accessible only by your API key) and Public (available in the community registry).
How long does custom model initialization take?
Most LoRA adapters load within 60 to 90 seconds. Full checkpoint deployments typically take 5 to 10 minutes depending on file size.
How does billing work?
LumozAI is pay-as-you-go. There are no subscriptions, monthly fees, or commitments. Load balance into your account and you are charged only for the model runs you make, via the same balance for the web UI and API.
What payment methods are supported?
We support all major debit/credit cards (Visa, MasterCard, Amex) through our secure payments processor.
Do credits expire?
Credits remain valid for 1 full year (365 days) from each reload date. Any fresh top-up extends your active credit window for another year.
What GPU hardware powers LumozAI?
Our fleet is powered by enterprise NVIDIA H100, A100, and L40S clusters interconnected with high-bandwidth NVLink and Infiniband networking to minimize inference step times.
Can I use webhooks for long-running video generations?
Yes. Simply include a webhook URL parameter in your POST /api/v4/run request. When generation finishes, our server sends an HTTP POST payload directly to your server with the output URL.
How is data security handled?
All traffic is strictly encrypted over TLS 1.3. We do not use your inputs or generation outputs to train any foundation models without explicit consent.