Limited-time offer: 20% off PixVerse V6 Max. Ends November 30. Try in Playground
Clear explanations of AI model APIs: what each kind of model does, how requests and billing work, and the parameters you send. Every page lists the live models it covers.
An LLM API lets your application send a conversation, as a list of messages and on many models with images, video or documents, to a large language model and receive its generated text. You choose the model per request, and usage is counted in input and output tokens.
An image generation API creates images from a text prompt, and on many models edits or combines images you supply. You send the prompt with settings such as aspect ratio and resolution, and the API returns the finished images as URLs or base64 data, billed per image.
A video generation API creates a video clip from a text prompt, and on many models from images, video or audio you supply. You submit a job, poll until the video is ready, and most models bill per second of output at a rate set by the resolution.
Image-to-video generation animates a still image into a short video clip. The image usually becomes the first frame, and your prompt describes the motion and the camera movement, so the subject, framing and style of the picture carry into the video.
A text-to-speech (TTS) API converts written text into spoken audio. You send the text, a voice and settings such as language and speed, and the API returns an audio file, or streams the audio so playback can start before the whole passage is ready.
A speech-to-text API transcribes spoken audio into written text. You upload a recording or pass its URL, or stream live audio, and the API returns the transcript, with timestamps, speaker labels or subtitle files on the models that support them.
A realtime voice API holds a live, two-way audio session over one WebSocket connection: your app streams microphone audio to the model and plays its spoken reply as it arrives. The user can interrupt the reply at any time, and the same endpoint also serves live transcription.
An AI music generation API composes audio from a text description of the style, mood, instruments and tempo, and on some models from lyrics, a reference recording or a video to score. You submit a job, poll until it finishes, and download the generated track.
A 3D generation API creates a 3D model, a textured mesh, from one or more images of an object, and on some models from a text prompt. You submit a job, poll until it completes, and download the asset in a format such as GLB, OBJ, FBX, USDZ or STL.
An embeddings API turns text, and on some models images or video, into vectors of numbers that place similar meanings close together. With a reranker to rescore the best matches, it forms the retrieval step of semantic search and retrieval-augmented generation.
An AI search API runs a live web search for an application and returns results shaped for language models: ranked pages with their text, a short answer with cited sources, or a multi-step research report. It gives an AI application information newer than a model's training data.
Tokens are the units a language model reads and writes: words, pieces of words, numbers or punctuation. Text models bill input and output tokens at separate rates per 1M tokens, while many media and search models bill per image, second, character, minute or request.
A context window is the maximum number of tokens a language model can handle in one request, counting everything it reads (instructions, conversation history, documents and tool definitions) plus the tokens it writes. A request that does not fit has to be shortened or split.
A reasoning model works through a problem in intermediate reasoning tokens before it writes its final answer. This improves results on multi-step tasks such as math, code and planning, uses more output tokens and time, and settings such as reasoning effort control it.
Temperature and top_p control how a model picks each next token: lower values make output focused and repeatable, higher values make it more varied. A seed fixes the random choices, so the same request with the same seed returns the same or a very similar result.
Structured output makes a language model return JSON that your code can parse instead of free text. JSON mode returns a syntactically valid JSON object, and JSON schema goes further by making the response match a schema you define, with its required fields and types.
Function calling, also called tool calling, lets a language model ask your application to run a function. You describe each function with a JSON Schema, the model replies with a function name and arguments, your code runs it, and the model uses the result to continue or answer.
Prompt caching reuses a model's earlier processing of the opening part of a prompt. When a request starts with the same tokens as a recent one, those cached tokens bill at a lower cache read rate on models that list one, and the rest bills at the normal input rate.
Batch inference runs many model requests as one asynchronous job instead of one call at a time. You upload a file of requests, they run within a completion window, and you download a file of results, which suits large jobs that do not need an immediate answer.
Check out our pricing or reach out if you want your own model deployed on our stack.