A video API is a web service that lets an application do something with video by sending requests instead of using a person and a desktop tool: store and play videos, broadcast live, run video calls, convert files, generate footage with AI, or edit and render finished videos.

Those are very different jobs, done by different companies, with different prices. "Which video API should I use?" only has an answer once you know which of the six jobs you need. This guide explains each one, names the providers that lead it in 2026, and shows how they charge. Prices were checked on each vendor's own pricing page on 18 October 2026.


The six types of video API

Type What it does Leading providers Usually billed by
Hosting and deliveryStores your videos and plays them anywhere (VOD)Mux, api.video, Cloudinary, Bunny Stream, Cloudflare StreamMinutes stored and minutes watched
Live streamingBroadcasts live video to many viewersMux, Amazon IVS, Cloudflare StreamHours or minutes in and out
Video calls (real time)Two-way video between people, or with AI agentsTwilio, Daily, LiveKit, Agora, VonageParticipant-minutes
Processing and transcodingConverts formats, resolutions and codecsAWS MediaConvert, Coconut, Cloudinary, Rendi, TransloaditOutput minutes or GB processed
AI video generationCreates new footage from a prompt or an imageGoogle Veo, Runway, Luma, KlingSeconds or clips generated
Editing and renderingAssembles finished videos from clips, images, text, voice and dataJSON2Video, Shotstack, Creatomate, Plainly, BannerbearMinutes rendered

Many products combine several: an app might generate clips with AI, edit them into a branded video with a voice-over, then host it for playback. The sections below take them one at a time.


1. Video hosting and delivery APIs

You upload a video once; the API encodes it, stores it and delivers it to any device through a player or a streaming URL (HLS). This is what you need for a course platform, a user-generated video feature, or any app where people watch videos that already exist.

ProviderPricingFree
Mux Video1080p: encoding free, storage $0.003 per minute a month, delivery $0.001 per minute watched100,000 delivered minutes a month; $20 monthly usage credit on pay as you go
api.videoEncoding free, storage from $0.00285 per minute, delivery from $0.0017 per minute watchedSandbox (30-second clips, watermarked, deleted after 24 h)
CloudinaryCredits shared across images and video: Plus $99 a month (225 credits)25 credits a month
Bunny StreamStorage from $0.01 per GB, delivery from $0.005 per GB, $1 monthly minimum14-day trial
Cloudflare Stream$5 per 1,000 minutes stored, $1 per 1,000 minutes delivered—

More detail: video hosting APIs.

2. Live streaming APIs

A broadcaster sends one live feed (often RTMP or WebRTC) and the API delivers it to many viewers with a few seconds of delay. Use it for live events, shopping streams and webinars where viewers watch rather than talk.

ProviderPricingFree
Mux LiveInput $0.03125 per minute, delivery $0.001 per minute watched, simulcast $0.02 per minute per targetPay-as-you-go credit
Amazon IVSInput from $0.20 an hour; output from $0.036 per viewer-hour (SD) to $0.144 (Full HD)12 months: 5 h input, 100 h SD output
Cloudflare StreamSame as on demand: per 1,000 minutes stored and delivered—

More detail: live streaming APIs.

3. Video call APIs (real time)

Two-way, sub-second video between participants: telehealth, tutoring, interviews, and increasingly voice and video AI agents that talk to users. The API handles WebRTC, servers around the world and recording.

ProviderPricingFree
Twilio Video$0.004 per participant-minute; recording $0.004, composition $0.01 per minute—
Daily$0.004 per participant-minute, falling to $0.0015 with volume; audio only $0.0009910,000 minutes a month
LiveKit CloudShip $50 a month (150,000 minutes, then $0.0005); open-source server you can self-host5,000 minutes + 1,000 agent minutes
AgoraPer 1,000 minutes: HD $3.99, Full HD $8.99, audio $0.9910,000 minutes a month
Vonage Video APIAbout $0.004 per participant-minuteTrial credit

More detail: video conferencing APIs.

Twilio announced the end of its video product in 2024 and reversed the decision months later; Twilio Video remains a standalone product, with a roadmap focused on healthcare, services and retail.

4. Video processing and transcoding APIs

Converts files you already have: change the format, codec, resolution or bitrate, extract audio or thumbnails, create HLS ladders. Hosting APIs do this internally; you need a separate processing API when you deliver from your own storage or need precise control.

ProviderPricingFree
AWS Elemental MediaConvertPer normalised output minute from $0.0075 (HD counts double, 4K four times)—
CoconutPer output minute: SD $0.0075, HD $0.015, UHD $0.03Test plan (1-minute inputs)
RendiFFmpeg commands as a service, by GB processed: Pro $25 a month for 250 GB50 GB a month
TransloaditLite $29 a month for 15 GB5 GB a month
Cloudinary1,000 transformations = 1 credit25 credits a month

If you would rather run it yourself, FFmpeg does all of this for free; our FFmpeg course shows how.

Full comparison, prices and FFmpeg-vs-JSON examples: video processing API guide.

5. AI video generation APIs

Generative models create new footage from a text prompt or a starting image. Clips are short (usually 4 to 10 seconds) and billed per second or per clip. They produce shots, not finished videos: to add a voice-over, captions, a logo or several scenes, you combine them with an editing API.

ProviderPricingClip length
Google Veo 3.1 (Gemini API, preview)Per second with audio: Standard $0.40, Fast $0.10, Lite from $0.05 (720p); 4K costs more4, 6 or 8 s; extendable
Runway APICredits at $0.01: Gen-4 Turbo $0.05 per second, Gen-4.5 $0.12 per second; also resells Veo and KlingModel-dependent
Luma API (Ray3.2)Per clip: 5 s from $0.15 (540p) to $1.20 (1080p)5 or 10 s
Kling APIPrepaid units; also available through Runway3 to 15 s

Note on OpenAI Sora: OpenAI removed the Sora video API on 24 September 2026. Lists that still recommend it are out of date.

Full comparison with clip lengths, audio, a 30-second cost example and how to assemble the clips: AI video generation API.

6. Video editing and rendering APIs

An editing API assembles a finished video from parts: your clips and images, text and titles, music, text-to-speech voice-overs, subtitles and transitions. You describe the video (in JSON or as a template with variables) and the API renders an MP4. This is the type used to make hundreds of videos from a spreadsheet, a product catalogue or an article feed.

ProviderHow you describe a videoEntry price
JSON2VideoScenes and elements in JSON; voice-overs and subtitles built in$49.95 a month for 200 min; free 10 min
ShotstackTimeline of tracks and clips in JSON$39 a month for 250 min
CreatomateEditor templates or RenderScript JSON$54 a month for 2,000 credits
PlainlyAfter Effects templates$69 a month for 50 min
BannerbearEditor templates and workflows (images first)$49 a month for 1,000 credits

Full comparison, examples and pricing per minute: Video editing API.

This is what JSON2Video does. Send a JSON with scenes, voice-over and subtitles, get an MP4 back, from code, n8n, Make or Claude Code. Try it free with 10 minutes of video →


Video APIs for AI agents (MCP)

Coding agents such as Claude Code, Cursor and Codex can now call APIs directly through the Model Context Protocol (MCP). An official MCP server means the agent can use the service without you writing glue code. Some vendors offer full servers that act on your account; others only expose their documentation.

TypeOfficial MCP server that acts on your accountDocumentation-only MCP
HostingMux, Cloudinary—
Video calls—LiveKit, Agora, Vonage
Processing—Rendi
AI generationRunway; Google's open-source Genmedia MCP (Veo)—
EditingJSON2Video, Creatomate, Plainly, Shotstack (beta)—

With an editing API connected, an agent can go from "make a 30-second vertical video about these three products" to a rendered MP4: it writes the video description, validates it, renders it and returns the link. See how it works with the JSON2Video MCP server, or how it compares with building videos in React with Remotion and Claude Code. For every way Claude can make videos, see can Claude make videos?


How to choose a video API

  1. Do the videos already exist? If people upload or record them, you need hosting (and maybe processing).
  2. Is it live? One-to-many: live streaming. Two-way conversation: video calls.
  3. Do you need footage that does not exist? AI generation, for shots and b-roll.
  4. Do you need finished videos made from your material or data? An editing API, often fed with clips from the other types.
  5. Then compare on the billing unit that matches your usage (minutes watched, participant-minutes, seconds generated or minutes rendered), the free tier, and whether it fits your stack: SDKs, no-code tools, and MCP if AI agents will use it.

Frequently asked questions

What is a video API?

A web service that lets software work with video through requests: hosting and playing videos, live streaming, video calls, converting files, generating footage with AI, or editing and rendering finished videos. Each of these is a different type of video API, usually from different providers.

What is the best video API?

It depends on the job. For hosting and playback, Mux and api.video are common developer choices; for live streaming, Mux or Amazon IVS; for video calls, Daily, LiveKit or Twilio; for processing, AWS MediaConvert or Coconut; for AI footage, Google Veo or Runway; for editing and rendering finished videos, JSON2Video, Shotstack or Creatomate.

Is there a free video API?

Several have free allowances: Mux (100,000 delivered minutes a month), Daily and Agora (10,000 call minutes a month), LiveKit (5,000 minutes), Cloudinary (25 credits a month), Amazon IVS (a 12-month free tier) and JSON2Video (600 credits, or 10 minutes of rendered video). AI generation APIs such as Veo are paid only.

Does OpenAI have a video API?

Not any more. OpenAI removed the Sora video API on 24 September 2026. For AI video generation by API, the main options are Google Veo (through the Gemini API), Runway, Luma and Kling.

What is the difference between a video API and a video SDK?

The API is the service you call over the internet; an SDK is a library for a given language or platform that makes calling it easier. Most video API providers offer both, and video call providers usually require their client SDKs in the browser or app.

Which video APIs work with Claude Code and other AI agents?

Through official MCP servers: Mux and Cloudinary for hosting, Runway for AI generation, and JSON2Video, Creatomate, Plainly and Shotstack (beta) for editing. LiveKit, Agora and Vonage offer documentation-only MCP servers.

Prices are list prices in US dollars from each provider's pricing page, checked on 18 October 2026. Vendors change prices often: confirm on their sites before deciding.

Published on April 12th, 2023

Author
Joaquim Cardona
Joaquim Cardona Senior Internet business executive with more than 20 years of broad experience in Internet business, media sector, digital marketing, online video and mobile technologies.