# Generate an illustration or photo from a prompt

> Generate an illustration or photo from a prompt: 1 companies fully do it, 13 with a limit, 42 carry the tag. Every listing in Image generation with the reason and a source.

- Canonical: https://www.anchorterminal.com/niches/image-generation.text-to-image/
- Markdown: https://www.anchorterminal.com/niches/image-generation.text-to-image/index.md (~1,550 tokens)
- Slim: https://www.anchorterminal.com/niches/image-generation.text-to-image/index.min.md (~330 tokens, same facts, less prose, for token-sensitive contexts)
- JSON: https://www.anchorterminal.com/niches/image-generation.text-to-image/index.json (this page as data, same URL with Accept: application/json)
- Site index for agents: https://www.anchorterminal.com/llms.txt (the key docs and every listing in one file, at most 16 MB, about 4 million tokens: https://www.anchorterminal.com/llms-full.txt)
- API: https://www.anchorterminal.com/api/v1/index.json
- MCP: https://www.anchorterminal.com/mcp (the whole catalogue as tools, no key)
- DNS: dig +short TXT <slug>.grade.anchorterminal.com (the grade as a DNS TXT record, for a sandbox with no HTTP where names still resolve; also top., best.<category>. and <a>.vs.<b>. under the zone, and the same under grade.letme.dev)
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)
- Updated: 2026-11-15

One company's docs show a full fit.

| Companies, full fit | With a limit | Carry the tag | Share of the category's work | First-hand accounts |
| --- | --- | --- | --- | --- |
| 1 | 13 | 42 | 12 | 6 |

## What counts as done

An image at the requested size and ratio that follows the prompt, saved by the agent, with the charge known.

Measured, for our own tests: the image is saved within 30 seconds of the call.

How agents ask for it: make an image of; illustrate this story segment; give me a hero image for this post; draw a cartoon cat

## Does the job

The docs show it meets the test above, and nothing we read says otherwise.

- **Recraft API (Recraft Inc.)**: Synchronous generation with a per-model size table (other WxH rejected), the charge in the credits field, about 1.1 seconds on V4.1 Flash V2, and url, b64_json or multipart results kept about 24 hours.

## Does the job, with a limit

It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.

- **Adobe Firefly API (Adobe Inc.)**: Aspect ratio and resolutionLevel enums set size, but every call is an async job to poll, no latency or per-image price is published, and access needs an enterprise contract.
- **Black Forest Labs FLUX API (Black Forest Labs Inc.)**: Aspect-ratio and resolution enums, a stated width and height range with 422 validation and a cost field cover size and charge, but no generation time is published against the 30 second bar, grounding web search is on by default, and results vanish after 10 minutes.
- **Bria API (Bria Artificial Intelligence Ltd.)**: Aspect ratios are listed and unsupported ones return 422, and sync: true returns the image in one call, but no generation time is published (4MP adds 30 seconds) and the charge isn't in the response.
- **fal image models (Features & Labels, Inc.)**: Price per image is on each model page and outputs stay on the CDN for at least 7 days, but sizes are per model and no latency is published per model against the 30 second bar; acc-35 ran stills cheaply on it.
- **FLORA (FLORA AND FAUNA FLOWERSHOP, INC.)**: Runs are async with estimated_seconds per model and a charged_cost on create, but the docs ask for about 15 seconds between polls and the REST API needs Starter or above.
- **Ideogram API (Ideogram AI, Inc.)**: Synchronous by default with size presets (others rejected) and an exact dry_run price, but no generation time is published against the 30 second bar and link lifetime is unstated.
- **Iris-3B (Speridlabs)**: About 0.7 seconds for 1024x1024 on one H100 at no charge, with --height and --width, but training was at about one megapixel in seven aspect buckets and the card points to an upscaler for larger prints; an NVIDIA CUDA GPU is required.
- **Leonardo.Ai API (Leonardo Interactive Pty Ltd)**: A sync generation endpoint returns the image in one call, with width and height enums per model and apiCreditCost, but cost arrives only after the call, credits aren't priced in a published table, and no latency is given against the 30 second bar.
- **OpenAI Image API (OpenAI OpCo, LLC)**: Size constraints are documented and usage tokens come back, but the guide says complex prompts may take up to 2 minutes, beyond the 30 second bar, and the charge must be computed from tokens.
- **Replicate image models (Replicate, LLC)**: Prefer: wait returns fast models in one call and official models carry a per-image price, but size rules are per model (acc-32), and no per-model latency is published against the 30 second bar; acc-16 ran FLUX-dev on it.
- **Seedream on BytePlus ModelArk (BytePlus Pte. Ltd.)**: Size rules (total pixel and ratio ranges) are stated per model, the response counts billed generated_images at a published price, and streaming returns each image as made, but no latency is published against the 30 second bar and a watermark is on by default.
- **Stability AI Image API (Stability AI Ltd)**: Generation is synchronous with aspect_ratio, seed and image bytes in the response, and failed results aren't charged, but no generation time is published against the 30 second bar.
- **SurgePix API (SurgePix)**: Illustration and poster skills return a PNG URL from a plain request (16:9 at 1536x864 for illustrations), but sizes are fixed per skill, credits are token-based with no per-call price, and task results 404 after 24 hours.

## Couldn't check

The docs were closed to us, or silent. These carry no mark either way.

- **Qwen-Image 2.1 (Model Studio) (Alibaba Cloud (Singapore) Private Limited)**: The card lists seven 2K aspect ratios, but the hosted API's size rules, latency and charge couldn't be read.

## What stops agents finishing

- explicit dimensions outside the model's range hard-fail
- price or API change with no local fallback
- result URL expires before download

## Around it

Before: llm.write-prompt. After: image-generation.edit-existing-image, video-generation.image-to-video, storage.upload.

## Where this comes from

Weight 12 of 100 in Image generation: 6 of 37 accounts; FLUX on Replicate for story visuals, Nano Banana replacing a stack, local FLUX to avoid price changes, Seedream size errors

As data: https://www.anchorterminal.com/api/v1/use-cases/image-generation.text-to-image.json, or the MCP tool get_use_case with id image-generation.text-to-image. Draft, read 2026-11-15 from public docs and first-hand accounts, not yet reviewed by a person.

---

Query this catalogue directly over MCP: https://www.anchorterminal.com/mcp (Streamable HTTP, no key). How: https://www.anchorterminal.com/agents/#mcp
