saymaker

Generate images and video on SayMaker from your own code: one API key runs Veo 3.1, Kling 3.0, Seedance 2.0, MiniMax H3, Nano Banana 2, GPT Image 2.5, Seedream 5.0 and the rest of the shelf on your own credits.

{:ok, url} = SayMaker.generate("a paper-cut layered mountain range at dusk")
{:ok, clip} = SayMaker.generate_video("a paper boat drifting down a rain gutter, low angle")

Each call waits for the run and returns the URL of the finished file. Runs on a free account wait in a queue before they start, and the wait grows with each run of the day; a paid plan starts at once. The default timeouts cover the longest wait.

Install

def deps do
  [{:saymaker, "~> 0.1"}]
end

Elixir 1.18+; uses :httpc and the standard JSON module, no dependencies. Set SAYMAKER_API_KEY or pass api_key:.

API key

Sign up on saymaker.ai (new accounts get sign-up credits), create a key at saymaker.ai/settings/apikeys, then pass it or export it:

export SAYMAKER_API_KEY=sk-...

A key runs on your account exactly like the site does: same models, same credit prices, same plan, and every run lands in your library at saymaker.ai/history.

Images

SayMaker.generate("isometric diorama of a ramen shop at night", aspect_ratio: "16:9")
SayMaker.generate("product shot of a glass perfume bottle on wet slate", model: "gpt-image-2-5-flare")

The default model is saymaker-image-v1, the cheapest text-to-image run. Pass a model id for nano-banana-2, gpt-image-2-5-flare, seedream-5-pro, qwen-image-3-pro and the rest; see the image models.

Editing a photo

Pass an image URL and the prompt describes the change rather than the picture:

SayMaker.generate(
  "change the jacket to dark green, keep the face, pose and background exactly as they are",
  image_url: "https://example.com/portrait.jpg"
)

Naming what must stay is what decides whether the edit holds. Prompts that only name the change tend to drift the whole picture.

Video

SayMaker.generate_video("a fox trotting through fresh snow at dawn, tracking shot", duration: 6)
SayMaker.generate_video("the camera slowly orbits the statue", image_url: "https://example.com/statue.jpg")
SayMaker.generate_video("a street drummer in the rain", model: "veo-3-1", resolution: "1080p")

The default model is minimax-h3-fast, the one a free account can run (480p or 768p, 4 to 15 seconds). Veo 3.1, Kling 3.0, Seedance 2.0 and the other video models need a plan or a credit pack; see pricing. Each model's options are on its page: Veo 3.1, Kling 3.0, Seedance 2.0.

Errors worth catching

Failures are {:error, reason, message}:

• :auth — no key, or the key is wrong or deleted
• :quota — the account is out of credits
• :plan — that model needs a paid plan or a credit pack
• :busy — another run on this account is still rendering; free accounts run one at a time
• :rejected — the content filter or the model refused the prompt; rewrite it rather than retrying

The API answers refusals with HTTP 200 and an error code in the body, so a client that only checks the status code reports a refusal as success. This one reads the body.

Free-account output carries a watermark; a paid plan returns the clean file.

MIT licensed. Made by SayMaker, the AI video generator agent.