API Models — Official API Examples

Runnable example code for the API Models unified API — one API key and one OpenAI‑compatible endpoint to call image, video, LLM, audio and embedding models (GPT‑Image, Gemini, Grok, Seedance, Kling, VEO, Claude, ElevenLabs and more), at up to 60–95% lower cost than calling each provider directly.

中文说明见 README.zh-CN.md

These examples use only the public API. There are no secrets in this repo — you bring your own API key, which you generate yourself on our website.

---

1. Get your API key

1. Sign up at apimodels.app (new accounts get $0.10 free credit).
2. Open Console → API Keys and create a key.
3. Export it as an environment variable (every example reads this):

export APIMODELS_API_KEY="sk-your-key-here"

Never hard‑code your key or commit it to git. Copy .env.example to .env for local use.

2. Base URL & auth

Base URL	https://api.apimodels.app/v1
Auth header	Authorization: Bearer $APIMODELS_API_KEY
LLM/chat	OpenAI‑compatible — point any OpenAI SDK at the base URL above

3. Quick start (cURL)

# Image — synchronous because `size` is WxH and there is no callback_url.
# Returns the OpenAI shape: { "created": ..., "data": [ { "url": "https://..." } ] }
curl -X POST https://api.apimodels.app/v1/images/generations \
  -H "Authorization: Bearer $APIMODELS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "model": "gpt-image-2-lite", "prompt": "a red panda barista, studio light", "size": "1024x1024", "response_format": "url" }'

4. What a $0.10 balance actually buys

New accounts start with $0.10. Every example in this repo except voice cloning runs inside that, and running all of them in this order still leaves change:

Example	Model	Cost per run	Fits in $0.10?
chat-*	claude-sonnet-4-6	< $0.001 (token‑based)	yes
embeddings.py	text-embedding-3-small	fractions of a cent	yes
text-to-speech.*	eleven-tts-flash	~$0.004 (min charge)	yes
image-generation.*	gpt-image-2-lite @ 1024×1024	$0.008	yes
image-edit.sh	gpt-image-2-lite @ 1536×1024	$0.015	yes (needs your own image URL)
video-generation.*	veo-3.1-fast-fhd	$0.07 flat	yes — but only this one video model
voice-clone.sh	MiniMax voice cloning	$15.80 once per voice	no — top up first

Two things worth knowing before you spend:

• Video is the expensive modality. veo-3.1-fast-fhd at $0.07 per clip is the only video model a fresh $0.10 account can complete. The next cheapest are grok-video-3 ($0.12 for its 6‑second minimum) and seedance-2.0-fast (~$0.36 for a short 480p clip) — both cost more than the whole signup credit and will return 402 Insufficient credits.
• Voice cloning is a flat $15.80, charged once when the clone succeeds. That is the upstream custom‑voice fee at cost, not a per‑second rate. Read the warning at the top of curl/voice-clone.sh before running it.

Failed generations are not charged. Live prices: apimodels.app/pricing.

5. What's in this repo

Folder	Language	Examples
curl/	Shell / cURL	image, image edit, video (+poll), chat, text‑to‑speech, voice clone, balance
python/	Python 3	image, video, chat (OpenAI SDK), embeddings, TTS, balance
node/	Node.js (ESM)	image, video, chat (OpenAI SDK), TTS, balance

6. Endpoints covered

Capability	Method & path	Notes
Image generation / edit	POST /images/generations	Async by default (data.taskId → poll GET /images/generations?task_id=). Pass a WxH size and no callback_url to get a synchronous OpenAI‑shaped reply instead. Pass image_url / image_urls to edit.
Video generation	POST /video/generations → GET /video/generations?task_id=	Async; poll for the result
Chat / LLM	POST /chat/completions	OpenAI‑compatible
Embeddings	POST /embeddings	OpenAI‑compatible
Text‑to‑speech	POST /tts/stream	Streams audio bytes. This is where the eleven-tts-* models live — they are not accepted by /audio/generations.
Other audio (music, dialogue, dubbing, MiniMax TTS)	POST /audio/generations → GET /audio/generations?task_id=	suno-v5, eleven-dialogue, eleven-isolator, eleven-dubbing, minimax-speech-*, kling-*
Account balance	GET /balance	Your remaining credits (USD); free, read‑only
Voices & cloning	GET /minimax/voices, POST /minimax/voice/clone	List, clone or design custom voices (cloning is $15.80 per voice)
Model list	GET /models	OpenAI‑compatible, no API key needed — the authoritative list of ids you can call

Full parameter reference: apimodels.app/docs.

7. Popular models

Every id below was checked against the live API. When in doubt, curl https://api.apimodels.app/v1/models is the source of truth.

• Image — gpt-image-2-lite, gpt-image-2, gemini-3-pro-image, grok-imagine-image, doubao-seedream-4-5-251128, kling-image-o1
• Video — veo-3.1-fast-fhd (cheapest, $0.07/clip), grok-video-3, seedance-2.0-fast, seedance-2.0, kling-v2-6, minimax-h3, veo-3.1
• LLM — Claude, GPT, Gemini, DeepSeek, Qwen, GLM families (OpenAI‑compatible names — see docs)
• Audio — eleven-tts-flash, eleven-tts-turbo, eleven-tts-multilingual, eleven-tts-v3 (all on /tts/stream); minimax-speech-02-turbo, suno-v5 (on /audio/generations)

8. Response shape

Task‑based endpoints return:

{ "code": 200, "msg": "success", "data": { /* ... */ } }

Async tasks return data.taskId; poll the matching GET ...?task_id=<id> until data.state is completed (then read data.resultUrls) or failed.

The OpenAI‑compatible paths (/chat/completions, /embeddings, /models, and /images/generations in synchronous mode) return OpenAI's own shapes instead.

Generated files are kept for 7 days. Download anything you want to keep — the result URL stops working after that.

License

MIT — use these examples freely in your own projects.

Prompt libraries

Curated prompts, each shown next to the clip or image it actually produced:

• Stunning MiniMax H3 prompts — 222 video prompts (repo) · browse with video
• GPT Image 2 prompts — 955 image prompts, browsable by category

Links

• Website: https://apimodels.app
• API Docs: https://apimodels.app/docs
• Pricing: https://apimodels.app/pricing
• Console (create API keys): https://apimodels.app/console/api-keys