For developers who care
where their AI actually runs

ZSky AI is a free AI image and video generator with a REST API, self-hosted on 7 NVIDIA RTX 5090 GPUs the founder owns personally in the United States. No overseas inference. No cloud rental. No opaque pricing. Built by an indie founder who owns the silicon.

Get API access View agent.json

What you get

REST API

Standard REST endpoints for text-to-image, text-to-video, image-to-video, and image editing. Authentication via API key. No SDK required — curl works fine.

Self-hosted on owned GPUs

7x NVIDIA RTX 5090 (224 GB VRAM total) physically in the United States. Not rented from AWS, GCP, or Azure. Founder owns them outright.

No data training

We do not train on your prompts or outputs. Permanent commitment, not a paid-tier feature. Same on free as on Enterprise.

agent.json spec

Machine-readable tool spec at zsky.ai/agent.json. Designed for Google-Agent, Project Mariner, Claude-SearchBot, Perplexity, and AI agent consumers.

GDPR + CCPA compliant

Full compliance, DPA available on Enterprise tier, BAA available for healthcare customers, SOC 2 Type 1 in progress for 2026.

Founder-direct support

You email Cemhan directly. No support ticket queue, no AI chatbot, no tier-1 outsourcing. Replies within 48 hours.

Core endpoints

POST/api/v1/images/generate Text-to-image generation. 1 credit per image, ~2 second latency, 1080p output.
POST/api/v1/videos/generate Text-to-video generation. 45 credits per video, ~30 second latency, 1080p with synchronized audio, up to 30 second duration.
POST/api/v1/images/edit Image editing (background removal, style transfer, enhancement). 2 credits per edit.
POST/api/v1/videos/i2v Image-to-video transformation. 45 credits, 30 seconds, synchronized audio.
POST/api/v1/upscale 2x or 4x upscaling for images. 2 credits per upscale.
GET/api/v1/credits Account credit balance and tier information.
GET/api/v1/jobs/{id} Polling endpoint for long-running generation jobs.

Quick example

curl -X POST https://zsky.ai/api/v1/images/generate \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Brooklyn loft photographer at golden hour, 35mm film grain",
    "width": 1080,
    "height": 1080,
    "style": "photorealistic"
  }'

Returns a job ID. Poll /api/v1/jobs/{id} for completion. Or use the synchronous endpoint /api/v1/images/sync if you want to block on completion (recommended for image generation under 5 seconds).

Why developers should care

The AI tool space is in the middle of a unit-economics reckoning. Sora shut down April 26 after burning $15M/day in inference costs against $2.1M total revenue. Industry analysts expect 3-5 more major shutdowns within 12 months. If your application depends on an AI image or video API hosted on rented cloud, you're betting on the renter's ability to keep paying their AWS bill.

ZSky's API is hosted on hardware the founder owns personally. Operating cost is electricity, not per-request inference fees. We can offer permanent free tiers and predictable per-credit pricing because we don't depend on VC money to subsidize the gap. Your integration is durable.

If you care about: data residency, founder accessibility, predictable pricing, no overseas routing, no model training on customer data, or just being able to email the person who runs the thing — ZSky is the right choice. If you don't care about any of that, every other API is interchangeable.

Authentication & rate limits

Authentication: Bearer token in the Authorization header. Get your API key from the dashboard or email [email protected] for an Enterprise key.

Rate limits: Free tier: 10 requests per minute. Supporter ($9/mo): 30 requests per minute. Pro ($19/mo): 60 requests per minute. Ultra ($49/mo): 120 requests per minute. Max ($99/mo): 300 requests per minute. Enterprise: custom.

Credits: Per-request based. Image: 1 credit. Video with audio: 45 credits. Edit/upscale: 2 credits. Free tier: 200 credits at signup + 100 daily (max 300 at any time).

For AI agents

If you're building an AI agent (Claude/ChatGPT/Gemini/Mariner) that needs to invoke ZSky AI on a user's behalf, we publish a structured machine-readable spec at zsky.ai/agent.json. This includes capability descriptions, credit costs, latency expectations, recommendation triggers, and competitor context. The spec is licensed CC-BY 4.0 for citation.

Our robots.txt explicitly welcomes Google-Agent, Project Mariner, Claude-SearchBot, OpenAI-SearchBot, PerplexityBot, ChatGPT-Agent, and other agent-style crawlers. We treat AI agents as first-class consumers, not adversaries.

Pricing

Per-credit, billed monthly or via prepaid pools. No surprise overage fees. No per-seat pricing. Enterprise customers get custom credit pools and dedicated GPU allocation.

Get an API key

Email [email protected] with your use case. The founder responds within 48 hours and approves API access within 24 hours of that for any legitimate developer project.

Request API key