Image and video generation API

One endpoint for image and video generation

GPT Image 2.5 and MiniMax H3 behind a single request body: model, prompt, parameters and reference images. Every job reports the exact cost it added to your balance.

Pay as you go · No subscription · Sign in with Google

POST /api/v1/generations

Example image generated from a text prompt
gpt-image-2.5image · 1024x1024
Example video frame generated from a text prompt
minimax-h3video · 768p frame
68f2c41b9a1e4d0012ab34cdsucceeded

"cost": 0.0400 · "result": { "images": [ 1 ] }

2
Models online
$0.0400
Per image call
1-15s
Video per job
9
Reference images max

Models

Two models, documented down to the parameter

Pick the model per request with a single field. The reference in the console is generated from the same definitions the API runs on.

gpt-image-2.5image

GPT Image 2.5

Text to image and image editing with up to 9 reference images in the same call.

  • Text to image
  • Image editing
  • Up to 9 references
  • Output size: 1024x1024, 1536x1024, 1024x1536, 1672x941, 941x1672 or auto

prompt: "A border collie and an old english sheepdog hosting a livestream."

Pricing $0.0400 per call
minimax-h3video

MiniMax H3

Text to video with optional reference images, up to 15 seconds per call.

  • Text to video
  • 480p to 1080p
  • Reference images
  • Duration: 1-15 seconds, 1080p accepts at most 10
  • Resolution: 480p, 768p or 1080p
  • Aspect ratio: landscape or portrait

prompt: "A paper boat drifts across a quiet pond at sunrise."

Pricing from $0.0400 per second

Capabilities

Assets that go straight to work

One account and one bill for every visual your team needs, from first draft to final cut.

Visuals from a written brief

Describe the shot you need and get a finished image for the campaign, the product page or the pitch deck, without booking a studio.

Edits that keep your subject

Keep the product or the person you already have and change everything around them: new background, new styling, new season.

Short-form video without a shoot

Turn a script line into a clip that is ready for the feed, framed the way your channel already publishes.

Make the assets you own move

Start from a still that is already approved and let it move, so product, composition and lighting stay on brand.

Built for volume, not one-offs

Queue the whole catalogue, every colourway and every market, and collect the results as they finish.

Results that drop into your stack

Every asset arrives as a plain link, ready to store in your CMS, catalogue or ad platform.

Quick start

First job in five minutes

  1. 1

    Sign in with Google

    One click creates your account. No passwords, no verification email.

  2. 2

    Create an API key

    The key is shown once when you create it and can be revoked from the console at any time.

  3. 3

    Send a job

    POST returns a job id, then read GET /api/v1/generations/{id} until it is succeeded or failed. Failed jobs are never charged.

curl -X POST $BASE_URL/api/v1/generations \
  -H 'Authorization: Bearer $API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "gpt-image-2.5",
    "prompt": "A border collie and an old english sheepdog hosting a livestream.",
    "parameters": { "aspectRatio": "1024x1024" },
    "images": []
  }'

Why HTopAPI

Metering you can check line by line

Pay only for what runs

Image jobs are billed per call, video per second of output. Failed jobs are stored with cost 0, so a retry costs nothing extra.

One request shape

Both models take the same body: model, prompt, parameters and reference images. One key covers every model on the API.

Prepaid balance, live cost

Your balance is the sum of top-ups minus the exact cost of every job, so the number in the console always matches your spend.

Usage you can audit

Each call is stored with its model, parameters and cost, and summarised per day and per model in the console.

Pricing

Pay as you go

Top up a prepaid balance and each job deducts exactly what it cost. Image calls are billed per call, video per second of output, and failed jobs are never charged.

  • No subscription, no seats, no monthly minimum
  • Failed jobs are never charged
  • Every charge visible in the console
  • Top-ups added by our team for now
Start building

FAQ

Questions, answered

Everything above comes from the model definitions and the metering the API actually runs on. Sign in for the full reference and a live cost estimate.

How do I get an API key?

Sign in with Google, open the console and create a key. The key is shown once when you create it, and you can revoke or replace it at any time.

What does one call cost?

GPT Image 2.5 is billed per call and MiniMax H3 per second of output: 480p at $0.0400/s, 768p at $0.0600/s and 1080p at $0.0800/s. The exact cost is returned in the response and recorded in the console.

How do I pay?

Pricing is pay as you go. You top up a prepaid balance and each job deducts exactly what it cost, so there is no subscription and no monthly minimum to cancel. Self-service top-up is not open yet, so contact us and we will add credit to your account.

How do I get a video result?

Video jobs are asynchronous. Submit the job, keep the returned id, then poll GET /api/v1/generations/{id} until the status is succeeded or failed. There is no webhook to configure.

What happens when a job fails?

A failed job is stored with cost 0, so nothing is deducted from your balance. Fix the request and submit it again.

Can I use the output commercially?

Yes. Assets generated through HTopAPI can be used in commercial products and campaigns. You are responsible for the prompts and reference images you submit.

Ship your first generation today

Sign in with Google, create a key, and send a job. Top up a balance when you are ready, and only pay for the jobs that succeed.