Blog
AI Image & Video API Guides
Hands-on model tests, production tutorials, prompt techniques, and current API pricing for generative media teams.
How to use gpt-image-2/text-to-image via the hiapi API: parameters, code, and a working example
gpt-image-2/text-to-image runs on the unified async /v1/tasks endpoint — POST to create, then poll GET /v1/tasks/{id} or use a callback. input takes prompt (required), aspect_ratio (e.g. 1:1, 16:9), and resolution (1K/2K/4K, defaults to 1K). Auth is Authorization: Bearer sk-<key>. A wrong key returns HTTP 200 with {"error":{"code":"permission_denied"}} — check body.error, not status. Final image lives at data.output[0].url. The URL expires (expireAt) — download immediately and re-host on your own storage.
How to Fix HiAPI Rate Limit (429) Errors
HiAPI returns HTTP 429 when a single API key sends too many concurrent or back-to-back requests to /v1/tasks or the polling endpoint. The most common trigger is fan-out: many parallel task creations from one key, or a tight polling loop on GET /v1/tasks/{taskId}. Fix it by adding exponential backoff (1s/2s/4s, cap 30s), capping client concurrency, switching from polling to callback.url, and slowing polling cadence to 2-5s with jitter. Check current usage at the HiAPI dashboard to confirm whether you are hitting account-level concurrency limits or running out of plan headroom. Use a Nano-Banana curl request to verify your key and network path once the retry layer is in place.
Z-Image API: Pricing, API Key, Examples, and When to Use It
Z-Image is Alibaba's Tongyi Z-Image, a fast, efficient photorealistic text-to-image model that also renders accurate in-image text in both English and Chinese. On HiAPI the callable model id is plain z-image. There is no separate z-image-turbo id to call, so don't code against one. Calls go through the unified async task API, POST /v1/tasks then poll until the image URL is ready. The input takes prompt plus aspect_ratio (and optional resolution). A size or width times height field is rejected, so leave it out. It's a low-cost, pay-as-you-go option. See live pricing rather than any number you read in a blog post.
Cheapest Image Generation API: How to Compare Real Cost Per Image
"Cheapest" is not a number you can look up once — it depends on resolution, model choice, and how many tries it takes to get a keeper. The metric that matters is cost per *usable* image: list price multiplied by attempts-to-acceptable, not the headline per-image rate. Retries and rework are the real budget killers. A model that needs five runs to land one good image is not cheap, even at a low sticker price. Iterate on a cheap, lightweight model at low resolution, then finalize the keeper on a flagship — you pay flagship rates only for images you actually ship. HiAPI is pay-as-you-go with one key across many image models, so you can route drafts and finals separately. Live numbers live on the pricing page; never trust a number copied into a blog post.
Image to Video API Workflow: Runway, Wan 2.7, Vidu, and Production API Patterns
Image-to-video is the right call when the opening frame is fixed — a product render, an approved key visual, a character sheet — and you need motion that respects it instead of a freshly hallucinated scene. {"The single biggest quality lever is the source image": "its resolution and aspect ratio cap the output, and a clean, well-framed subject gives the model far less room to drift."} First-frame control pins where the clip starts; adding a last frame pins where it ends, which is the most reliable way to kill end-of-clip motion drift. Prompt the change, not the scene. With i2v the model can already see the subject, so the prompt's job is the motion and the camera, not a description of what's in the frame. On HiAPI the callable workflow runs on Seedance 2.0 and Wan 2.7 image-to-video through one async task API — create, poll or get a callback, download — billed per second, so resolution and duration are your cost.
GPT Image API Guide: GPT-4o, GPT-Image-1, GPT Image 2, and Migration Notes
OpenAI's image line is one evolving model, not four parallel ones: GPT-4o image gen (2025) became the gpt-image-1 API, then gpt-image-1.5, and now GPT Image 2 (April 2026), with each generation replacing the previous default. The current HiAPI text-to-image model id is gpt-image-2/text-to-image. GPT Image 2 is the reasoning-powered current generation, strongest at following instructions and rendering legible text on the image. On HiAPI the GPT image family available is GPT Image 2, including text-to-image, image-to-image, and multiple routes; gpt-image-1 and gpt-image-1.5 are not callable models. Migration uses the unified async API: submit with POST /v1/tasks, then poll GET /v1/tasks/:id or configure a callback.
Free AI Image Generation API Without an API Key: What Actually Works
A truly key-less, production-grade image API does not exist — abuse and GPU cost make it impossible to run one without authentication. Real no-key options (Pollinations) and free tiers (Cloudflare Workers AI, Hugging Face, Google AI Studio) work for throwaway and hobby use, but each trades away watermark-free output, rate headroom, commercial rights, or stability. Free-trial providers like fal.ai and Replicate still require an API key — the free part is starter credit, not a key-less endpoint. For reliable or commercial work, the practical path is a no-card signup that grants free credit to test, then pay-as-you-go — one key, many models. HiAPI gives $1 free credit (currently ~50 images) to try real generations across 13 image models via one unified task API; production calls authenticate with a key.
Free Text to Video API: Trials, No-Key Demos, and the Production Reality
A truly key-less text-to-video API you can script against does not exist — video is too compute-heavy for anyone to leave an open, unauthenticated endpoint running. No key in the wild means one of two things: consumer web tools you click through in a browser (not an API), or free-credit API tiers that still hand you a token first. {"Free tiers are real but capped": "short clips, watermarks, 720p ceilings, daily credit refreshes, queues, and unstable policies that change month to month."} For production you authenticate every call. The low-friction path is a no-card signup, free credit to test, then pay-as-you-go. On HiAPI, video runs through the async task API (Seedance 2.0, Wan 2.7), billed per second; signup gives $1 free credit (currently ~50 images of credit) to test before you pay.
Fal.ai API Alternative: Keys, Docs, Pricing, and Migration Notes
fal.ai is a developer-first generative-media platform with serverless GPU and pay-per-use billing; if it fits your stack, it is a solid choice. HiAPI is a fair alternative when you want one unified API across image and video models, a single key, and pay-as-you-go pricing. HiAPI generation uses the unified async /v1/tasks API; existing clients must map model names, input fields, and result handling to that task contract. Migration means getting a HiAPI key, submitting a task, then polling or receiving a callback before downloading the result. Pricing on both is output-based; compare the model and read HiAPI's live pricing instead of relying on frozen figures.
Invalid API Key Errors: Causes, Fixes, and an API Key Setup Checklist
Most invalid-key errors are not bad keys — they're whitespace, a wrong env var, a stale key, or a key sent to the wrong base URL. Read the error body, not just the status: it names the real cause and usually carries a request_id you quote to support. On HiAPI an auth failure is HTTP 401 with code permission_denied and type hiapi_error — not error_key_does_not_exist or invalid_api_key (those are other ecosystems' strings). claude code invalid api key is almost always the Anthropic key or a base-URL override — fix that, not HiAPI. Run the 8-step checklist below to validate any key with one minimal API call before you debug your app.
Build with the models you just read about.
Compare image, video, and audio models, then use the same API workflow in production.