Cheapest VLM Caption API

A production REST API that analyzes any image and returns a natural language caption, an array of SEO tags, and a WCAG-compliant alt text string — all in one HTTP call. $0.0004 per image — over 60% cheaper than AWS Rekognition ($0.001) and over 73% cheaper than Google Cloud Vision ($0.0015). The cheapest VLM caption API available. 5,000 free credits on a 24-hour trial, no credit card required.

Quick start — one API call

Sign up, copy your key from the dashboard, and POST your image. Choose a mode and style, then poll the generation endpoint until status=completed to read the caption, tags, and alt text.

curl -X POST https://api.pixelapi.dev/v1/image/caption \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "image=@product.jpg" \
  -F "mode=full" \
  -F "style=product" \
  -F "max_tags=15"
# Response: {"generation_id": "uuid", "status": "queued", "credits_used": 0.0004}

# Poll until completed
curl https://api.pixelapi.dev/v1/image/UUID \
  -H "Authorization: Bearer YOUR_API_KEY"
# Response: {
#   "status": "completed",
#   "caption": "A white ceramic coffee mug with a minimal handle on a natural wood table.",
#   "tags": ["coffee mug", "ceramic", "white", "minimalist", "kitchen"],
#   "alt_text": "White ceramic coffee mug on wooden table."
# }

pip install pixelapi
---
from pixelapi import PixelAPI

client = PixelAPI(api_key="YOUR_API_KEY")
result = client.caption(
    image="product.jpg",
    mode="full",       # full | caption | tags | alt_text
    style="product",   # product | creative | seo | technical
    max_tags=15,
)
print(result.caption)   # natural language description
print(result.tags)      # ["coffee mug", "ceramic", ...]
print(result.alt_text)  # "White ceramic coffee mug on wooden table."

npm install pixelapi
---
import { PixelAPI } from "pixelapi";

const client = new PixelAPI({ apiKey: process.env.PIXELAPI_KEY });
const result = await client.caption({
  image: "./product.jpg",
  mode: "full",      // full | caption | tags | alt_text
  style: "seo",      // product | creative | seo | technical
  maxTags: 15,
});
console.log(result.caption);   // natural language description
console.log(result.tags);      // string[]
console.log(result.altText);   // WCAG alt text string

composer require pixelapi/pixelapi
---
<?php
use PixelAPI\Client;

$client = new Client(getenv("PIXELAPI_KEY"));
$result = $client->caption([
    "image"    => "product.jpg",
    "mode"     => "full",
    "style"    => "product",
    "max_tags" => 15,
]);
echo $result->caption;
print_r($result->tags);
echo $result->alt_text;

gem install pixelapi
---
require "pixelapi"

client = PixelAPI::Client.new(api_key: ENV["PIXELAPI_KEY"])
result = client.caption(
  image:    "product.jpg",
  mode:     "full",
  style:    "product",
  max_tags: 15,
)
puts result.caption
puts result.tags.inspect
puts result.alt_text

go get github.com/pixelapi/pixelapi-go
---
import "github.com/pixelapi/pixelapi-go"

client := pixelapi.New("YOUR_API_KEY")
result, err := client.Caption(pixelapi.CaptionParams{
    Image:   "product.jpg",
    Mode:    "full",
    Style:   "product",
    MaxTags: 15,
})
if err != nil { panic(err) }
fmt.Println(result.Caption)
fmt.Println(result.Tags)
fmt.Println(result.AltText)

Pricing — cheapest VLM caption API vs AWS, Google, Azure

AWS Rekognition pricing verified from aws.amazon.com/rekognition/pricing, October 2026. Google Cloud Vision pricing verified from cloud.google.com/vision/pricing, October 2026. Azure and OpenAI pricing pages could not be machine-read at time of publication — visit their respective pricing pages for current rates. PixelAPI's per-image price is set at half the cheapest verifiable mainstream rival per our pricing principle.

What you get back

The caption API is a pure text API — it returns structured JSON, not images. Each response contains up to three fields depending on the mode you select.

Caption modes and styles reference

Common workflows

The caption API is purpose-built for these production automation patterns:

Integrations & SDKs

The Caption API is a standard REST endpoint compatible with any HTTP client or automation platform. Official SDKs ship retries, polling, and binary handling out of the box.

Comparison vs alternatives

How PixelAPI's caption API compares to the mainstream vision APIs on price, output format, and developer experience:

Rate limits & error handling

Rate limits apply per API key. Exceeding the limit returns HTTP 429 with a Retry-After header indicating seconds to wait before retrying.

The Python and Node SDKs handle 429 retries with exponential backoff automatically. For manual implementations, start at 2 seconds, double on each retry, and cap at 30 seconds.

# Python SDK — auto-retries 429 with backoff
from pixelapi import PixelAPI
client = PixelAPI(api_key="...", max_retries=4)
result = client.caption(image="product.jpg", mode="full")
# Retries automatically on 429; raises after max_retries exceeded

If an image is rejected (corrupt file, unsupported format, NSFW content flagged at the gate), the job returns a 400 or 422 error and credits are not deducted. Failed jobs that pass validation but error during processing are automatically detected and refunded.

Frequently asked questions