Glossary

A

API key

A secret string that identifies your account when your code calls an API. GotoRamp keys are sent as a bearer token, and each key has its own usage and spend limit; keys open with early access.

Aspect ratio

The shape of a video frame, written as width to height. The Seedance 2.0 line supports 21:9, 16:9, 4:3, 1:1, 3:4 and 9:16, plus adaptive, which follows the first frame. See the video formats guide.

Async job (task ID)

A request that returns at once with an ID and finishes later. Seedance video on GotoRamp works this way: POST /v1/video/generations returns a task_id, and GET /v1/video/generations/{task_id} reports progress.

B

Backoff

Waiting progressively longer between retries after a rate-limit or temporary error, usually with a small random delay (jitter) added. It keeps many clients from retrying at the same moment. See the OpenAI SDK guide for an example.

C

Chat completion

A request in which you send a list of messages (system, user and assistant turns) and the model returns the next message. On GotoRamp it's POST /v1/chat/completions.

Code-switching

Mixing two or more languages in one sentence or conversation, as in Singlish, Manglish and Taglish. Include it when you test models; see choosing a model for Southeast Asian languages.

Concurrency

The number of requests or jobs running at the same time. Each account has a concurrency limit; the Volume plan has higher concurrency, and Enterprise can reserve capacity.

Context window

The maximum number of tokens a model can handle in one request, counting both input and output. It varies by model version; check the console or GET /v1/models.

D

Data processing agreement (DPA)

A contract that sets how a service provider handles personal data on your behalf. GotoRamp offers a DPA on the Enterprise plan; see pricing.

E

Early access

GotoRamp's current stage: no customer accounts are live yet, and the site collects early-access requests. API keys and the endpoint open with your invitation, and rates are shared with early-access accounts. Request early access.

F

First frame / last frame

Images that fix how a video clip starts and, optionally, how it ends. On Seedance a last frame needs a first frame; use them to keep a product accurate or to land on a designed end card.

I

Image-to-video

Generating a video that starts from an image you supply, such as a product photo. The prompt then describes the motion rather than the picture.

Input and output tokens

Input tokens are what you send: the system prompt, conversation history and any documents. Output tokens are what the model writes back. GotoRamp bills them at separate rates, per model.

J

JSON output

A setting or instruction that makes a model reply in valid JSON, sometimes matched to a schema you provide. Support varies by model version, so validate every response in your code.

M

Model ID

The exact string you pass in the model field, such as deepseek-chat or seedance-2.0. GET /v1/models lists the IDs your account can call.

N

Native audio

Sound generated together with the picture in the same video job, instead of added afterward. On the Seedance 2.0 line you can turn it on or off per request.

O

OpenAI-compatible API

An API that accepts the request and response format the OpenAI SDKs use, so existing code works after you change the base URL and key. GotoRamp's text API is OpenAI-compatible; see the OpenAI SDK guide.

P

PDPA

Short for Personal Data Protection Act, the name of the data protection law in Singapore (2012), Malaysia (2010) and Thailand (B.E. 2562 / 2019). The Philippines has the Data Privacy Act of 2012. See the Singapore, Malaysia and Thailand pages.

Per-second billing

Charging for video by the length of what is returned. GotoRamp bills Seedance per second of video returned, at a rate set per model and resolution; failed requests aren't billed.

Per-token billing

Charging for text by the number of tokens processed. GotoRamp bills text models per token, with input and output tokens at separate rates for each model.

Polling

Checking the status of an async job repeatedly until it finishes. Poll at growing intervals rather than in a tight loop, and stop after a sensible timeout.

Prepaid balance

Money you add to your account before you use the API; usage is deducted from it. GotoRamp works from a prepaid balance in US dollars, with no subscription, seat fee or monthly minimum.

Provenance / AI label

Information that shows content was made or changed with AI, such as a visible label, a caption or embedded metadata. Many platforms and some jurisdictions require disclosure of realistic AI content; see the publishing checklist.

R

Rate limit

A cap on how many requests you can send in a period, or have running at once. Going over it returns HTTP 429; wait and retry with backoff.

Reasoning model

A model that works through intermediate steps before it answers. Reasoning models suit multi-step problems and are usually slower, often producing more output tokens. The DeepSeek and Qwen families include reasoning models.

Reference-to-video

Video generation steered by reference images, video or audio that set identity, motion or rhythm without fixing the first frame. On seedance-2.0 a job can carry up to 9 images, 3 video clips and 3 audio clips.

Resolution

The pixel size of each video frame, named by its shorter side. The Seedance 2.0 line outputs 480p or 720p; use 480p for drafts and 720p for finals.

S

Seed

A number that sets the starting point of random generation. With the same seed and the same inputs, a video result repeats more closely, which helps you see the effect of a prompt change.

Spend limit

A cap on how much one API key can spend. GotoRamp sets spend limits per key; once a key reaches its limit, requests on that key are refused until you raise it.

Streaming

Receiving a model's reply in pieces as it's generated instead of all at the end. On GotoRamp, set "stream": true on a chat completion request.

T

Text-to-video

Generating a video from a text prompt alone. See how to write Seedance video prompts.

Token

The unit of text a model reads and writes: a word, part of a word, a character or a punctuation mark, depending on the tokenizer. Text models on GotoRamp are billed per token.

Tokenizer

The part of a model that splits text into tokens. Each model family has its own, so the same text can produce different token counts in different models and scripts; measure with the usage field of each response.

Tool calling

A model's ability to return a structured request to run a function you defined, with its name and JSON arguments; your code runs it and sends back the result. Also called function calling. Support varies by model version.

V

Vision-language model

A model that takes images as well as text as input and answers in text, for example to read a receipt or describe a product photo. The Qwen family includes vision-language models.