Sulat.com / Writing on AISunday, November 15, 2026
Developer Tools/15 November 2026

Cap opens its video pipeline as an API, with free uploads and encoding

Cap Infra, announced on 13 November by Cap founder Richie McIlroy, charges US$0.85 per 1,000 minutes of 1080p storage and US$0.95 for delivery. The price comparison with Mux and Cloudflare Stream is Cap's own.

Glass 3D Cap logo floating in a bright GPU server room, with blue light streams and translucent video frames flowing from the racks.

At a glance

  • Cap, the open-source screen recorder, launched Cap Infra on 13 November, a video and speech API built on the GPU encoding pipeline that processes every Cap recording, according to founder Richie McIlroy's launch post.
  • An upload returns adaptive HLS streaming from 360p to 4K, an MP4, a poster frame, a hosted player and an iframe embed. Cap says uploading, encoding and the player are free.
  • Storage costs US$0.85 per 1,000 minutes a month up to 1080p, US$1.45 at 1440p and US$2.80 at 4K. Delivery is US$0.95 per 1,000 minutes and speech to text is US$0.10 per audio hour.
  • Cap's own table puts 10,000 minutes of 1080p storage at US$8.50 against US$30 on Mux and US$50 on Cloudflare Stream; the table leaves out Mux's 100,000 free delivery minutes each month.

Cap, the open-source screen recorder, has opened the video pipeline behind its app to outside developers as Cap Infra, a video and speech API. Cap says uploading and encoding are free, storage starts at US$0.85 per 1,000 minutes a month and delivery costs US$0.95 per 1,000 minutes.

Cap founder Richie McIlroy announced it in a launch post on 13 November. “The infrastructure we built to run Cap is now open to everyone,” the post says. Cap’s account shared the link, and McIlroy called it “the infra behind Cap” that is “now for everyone”.

What Cap Infra does

Every Cap recording is uploaded, encoded on Cap’s own GPU farm, transcribed and served from a CDN, usually within seconds of pressing stop, Cap says. Cap Infra gives developers the same path through a REST API. A developer uploads a video and gets back adaptive HLS streaming (the format that lets a player switch quality as a viewer’s connection changes) in up to five renditions from 360p to 4K, never above the source resolution. The upload also returns an MP4, a poster frame, a hosted player page and an iframe embed.

Cap Infra flow: a video upload is encoded and returns an adaptive stream, MP4, poster frame, hosted player page and iframe embed.
One upload returns a stream, a file, a poster, a player page and an embed.

Speech to text works on its own. A developer sends a public URL or uploads an audio or video file and gets a transcript with word and segment timings. The video doesn’t have to be hosted on Cap. Cap says an hour of audio takes under 24 seconds on its GPUs in its own tests.

How an upload works

A developer’s server calls POST /api/developer/v1/uploads with a secret key that starts with csk_. Cap’s getting-started docs say the key stays on the server. Cap returns a video ID and an upload token that starts with cut_ and works for one hour by default. The browser sends the file in parts straight to storage with that token, and the server polls a status endpoint until the video is ready or errored.

Two options limit what a token can do. corsOrigin restricts uploads to one site’s origin, and maxBytes caps the file size. Without maxBytes, the docs say, a token can upload a file of up to 36 GB into storage the developer pays for.

Videos are encoded up to 1080p by default, and 1440p and 4K are set per upload. A ready video has HLS and MP4 URLs that need no API key, so any player can use them. A secret key allows 300 requests a minute across every endpoint, and an over-limit call gets a 429 reply with a Retry-After header.

Cap says a video plays as soon as its first output is done: under five seconds after the upload finishes for an hour of 4K, with every rendition following in about two minutes. Those are figures from Cap’s own tests.

Cap Infra also has a hosted MCP server at cap.so/api/developer/mcp that takes the same secret key. Its tools create and complete uploads, wait for a video, list and delete videos, create and read transcripts, and check usage. The docs show setup for Claude Code, Cursor and Codex, and Cap’s Infra page offers a single prompt that a coding agent can follow to wire up uploads and playback.

What it costs

Cap charges for stored video, delivered video and transcription. Uploading, encoding and the hosted player are free.

MeterUp to 1080p1440p4K
Uploading and encodingFreeFreeFree
Storage, per 1,000 minutes a monthUS$0.85US$1.45US$2.80
Delivery, per 1,000 delivered minutesUS$0.95US$0.95US$0.95
Speech to text, per audio hourUS$0.10US$0.10US$0.10
Player page and iframe embedIncludedIncludedIncluded
Cap Infra prices, from Cap’s launch post of 13 November 2026.
Cap Infra pricing card: uploading, encoding and the player are free; storage US$0.85, delivery US$0.95 per 1,000 minutes, speech to text US$0.10 per audio hour.
Uploading and encoding cost nothing; storage, delivery and transcripts are billed.

Billing runs on prepaid credit. Each app has its own balance, auto top-up is optional, and there are no seats or plans. An app with no credit can’t upload. A video is charged one month of storage when it first becomes ready, then daily after that month. The docs give an example: a 10-minute 1080p video costs US$0.0085 when it becomes ready, then US$0.00028 a day after its first month.

Delivery is billed from 14 November 2026, the launch post says, and a delivered minute is a minute of video that viewers’ players download. Cap’s Infra page lists storage and speech to text as its two charges and doesn’t list delivery. Cap Pro, the paid tier of the Cap app, includes US$5 of Cap Infra credit a month, which Cap says covers 5,800 minutes stored at 1080p for 30 days or 50 hours of transcription.

How Cap compares itself with Mux and Cloudflare Stream

Cap’s launch post includes a table of list prices taken from each provider’s pricing page on 13 November, before volume discounts and free allowances.

WhatCap InfraMuxCloudflare Stream
EncodingFreeFree on basicFree
Storing 1080p, 10,000 minutes for a monthUS$8.50US$30.00US$50.00
Storing 4K, 1,000 minutes for a monthUS$2.80US$9.60US$5.00
Delivering 1080p, 1,000 delivered minutesUS$0.95US$1.00US$1.00
Delivering 4K, 1,000 delivered minutesUS$0.95US$3.20US$1.00
Cap’s own comparison of list prices, as published in its launch post.
Bar chart of storing 10,000 minutes of 1080p video for a month: Cap Infra US$8.50, Mux US$30.00, Cloudflare Stream US$50.00, from Cap's own table.
On Cap’s own list-price table, a month of 1080p storage costs a fraction of its rivals’.

Mux’s pricing page matches the 1080p storage line: it lists US$0.003 per minute a month for 1080p basic quality, which is US$30 for 10,000 minutes. It also lists delivery at US$0.001 per minute after 100,000 free delivery minutes each month. That free allowance isn’t in Cap’s table. The Mux 4K and Cloudflare Stream figures in the table are Cap’s.

Compliance and licence

Cap says it’s HIPAA compliant and holds SOC 2 Type II and ISO 27001 certifications, and that Cap Infra runs on the same infrastructure. Organisations can buy a signed business associate agreement, the contract HIPAA requires when a vendor handles health data, from their organisation settings in the Cap dashboard.

The API, media server and player are on GitHub in CapSoftware/Cap. Cap’s product reference says most of the repository is licensed under AGPLv3, with commercial terms described at cap.so/docs/commercial-license.

The speed and transcription figures come from Cap’s own tests, and the launch post doesn’t describe the test setup. Developers can create an app and a secret key in the Cap Infra console, with the API reference in the docs.


Discover more from Sulat AI & Tech News

Subscribe to get the latest posts sent to your email.