Image descriptions for alt text and creative prompts.

Give a vision model an image and a question. Get a description of the subject, composition, lighting, colors, or visible text, then copy the result into your next brief.

Explore Describe No subscription. Sign in to create.

What does $10 make?

Compare a stated token budget before choosing a vision model.

$10$500
357example analysis runs

$0.028 / reference token workload

  • Gemini 3.1 Pro
Example: 1,000 total input tokens and 1,000 output tokens, including any image tokens within that input allowance. This is not a fixed price per image: image processing and response length change actual usage. There is no minimum text job charge.

More useful than a generic caption.

Ask about what is visible

Explore subjects, layout, lighting, color, style, and readable text in a single source image.

Choose the question and model

Use the starting visual-analysis prompt or write your own. Switch between the vision models enabled in SUPERMODELS.

Take the words into your workflow

Copy the response into a brief, revise it into alt text, or use it to develop your next image prompt. Check important details against the source.

10 vision models to explore.

Image-capable text models available in Describe.
Google

Gemini 3.1 Pro

Gemini 3.1 Pro brings multimodal reasoning into a writing tool with an explicit output-token budget.

$0.028 / reference runImage input · 1,000 input + 1,000 output tokens
Alibaba

Qwen3.7 Plus

Qwen3.7 Plus combines text reasoning with image inputs in the tool.

$0.001928 / reference runImage input · 1,000 input + 1,000 output tokens
OpenAI

GPT-5.6 Sol

GPT-5.6 Sol is configured here for text and image-led reasoning with adjustable response length.

$0.024 / reference runImage input · 1,000 input + 1,000 output tokens
Anthropic

Claude Opus 5.5

Claude Opus 5.5 provides a configurable writing tool for drafting, analysis, and technical work.

$0.048 / reference runImage input · 1,000 input + 1,000 output tokens
Anthropic

Claude Fable 5

Claude Fable 5 offers a text-and-image studio with a substantial output budget for developed responses.

$0.096 / reference runImage input · 1,000 input + 1,000 output tokens
Anthropic

Claude Sonnet 5

Claude Sonnet 5 provides a configurable studio for writing, analysis, and code with optional image context.

$0.024 / reference runImage input · 1,000 input + 1,000 output tokens
Google

Gemini 3.5 Flash

Gemini 3.5 Flash offers multimodal text generation with adjustable temperature and output length.

$0.021 / reference runImage input · 1,000 input + 1,000 output tokens
OpenAI

GPT-5.6 Terra

GPT-5.6 Terra is the middle GPT-5.6 option in this catalog, with text and image inputs and a configurable response budget.

$0.035 / reference runImage input · 1,000 input + 1,000 output tokens
OpenAI

GPT-5.6 Luna

GPT-5.6 Luna is the lower-cost GPT-5.6 option in this catalog for compact writing and analysis tasks.

$0.0098 / reference runImage input · 1,000 input + 1,000 output tokens
Anthropic

Claude 4.5 Haiku

Claude 4.5 Haiku provides a compact text-and-image workflow with adjustable response length.

$0.0084 / reference runImage input · 1,000 input + 1,000 output tokens

Describe images

Prepare your inputs and inspect the estimate. Generation requires an account and wallet credit.

Describe an image

PROMPT209 characters
IMAGE TO TEXT

Edit the starting prompt for a precise analysis, or leave it as-is for a comprehensive description.

A few useful answers.

What can an image describer help with?

It can draft a visual description, identify the main subject and setting, discuss composition and lighting, or extract clearly visible text. You can ask for concise alt-text suggestions or a detailed creative brief, then review and edit the answer.

Does Describe recreate or edit the image?

Describe returns text about your source image. To create a new image or edit a reference, take your description into the image creation tool and choose a model with the inputs you need.

Is there a fixed price per image description?

No. Describe uses the selected text model’s input and output token pricing. Image processing, prompt length, and the generated response affect usage. The calculator shows a stated example token workload; the platform shows its prompt and output-budget estimate.

Can I rely on every detail in the response?

Review descriptions against the original image, especially small text, quantities, identities, and subtle details. AI descriptions can be incomplete or incorrect. Edit suggested alt text for the image’s purpose and surrounding content.

Make something.

Free sign up

No subscription. Add credit when you’re ready to create.