# Highlander

> Highlander is a realtime AI video generation API powered by MiniMax Fast H3, producing 1344x768 clips at 24 fps with native synchronized audio at $0.02 per second of output.

Highlander is a realtime AI video generation API built on MiniMax Fast H3, serving efficient video streams from a single 8xH100 node in Palo Alto, CA. It targets developers who want programmatic access to high-quality video generation without subscriptions, seat minimums, or sales calls. The service is live and publicly accessible, with $1.00 of free credit on signup and no credit card required to start.

## What It Is

Highlander is a usage-based video generation API that wraps the FastH3 VSA checkpoint — a distilled version of MiniMax H3 — and exposes it through a simple HTTP interface at `api.highlander.sh`. Every request produces a 1344×768 MP4 at 24 fps with a native synchronized audio track generated in the same model pass. The service positions itself as the cheapest realtime video option at its quality level, with a single model and a single rate rather than a tiered quality ladder.

## How the Pricing Model Works

Highlander charges per second of generated video output, not per seat or per month. The default schedule uses 5 diffusion steps; raising the step count scales the price proportionally because steps are where the compute is consumed. Failed, timed-out, or worker-rejected generations are automatically refunded — users are never charged for a clip they did not receive. Credit is purchased as one-off prepaid payments through Stripe and never expires.

## Output Specs and Performance

- **Resolution:** 1344×768
- **Frame rate:** 24 fps
- **Audio:** Native synchronized audio generated in the same model pass as video
- **Render speed:** Approximately 0.94× realtime end-to-end at the default 5-step schedule
- **Clip lengths:** 90 frames (~3.75 s) up to 345 frames (~14.375 s) at the default schedule
- **Infrastructure:** Single 8xH100 node, one job at a time

The homepage displays unedited MiniMax Fast H3 output transcoded for the web, covering cinematic scenarios such as city renders, motorcycle chases, and FPV canyon runs.

## Deployment Model and Capacity

The current deployment is a single 8xH100 node serving one job at a time. The pricing page notes that dedicated capacity or concurrency requires direct contact before building against it. The prepaid credit model is also framed as a security feature: a leaked API key can never spend more than the balance behind it, and no unexpected invoices are generated.

## Who It Is For

Highlander targets developers and technical users who want API-level access to realtime video generation for prototyping, production pipelines, or experimentation. The playground at `highlander.sh/playground` allows prompt-based generation without writing code, while the docs at `highlander.sh/docs` provide the full request/response reference for API integration.

## Features
- Realtime AI video generation via API
- MiniMax Fast H3 (FastH3 VSA checkpoint) model
- 1344x768 resolution output
- 24 fps native video
- Synchronized audio generated in the same model pass
- Prepaid credit, no subscription
- Automatic refund on failed generations
- Configurable diffusion step count
- Web playground for prompt-based generation
- Stripe-based one-off credit purchases
- $1.00 free credit on signup, no card required

## Integrations
Stripe, MiniMax H3

## Platforms
API, WEB

## Pricing
Paid

## Links
- Website: https://www.highlander.sh
- Documentation: https://www.highlander.sh/docs
- EveryDev.ai: https://www.everydev.ai/tools/highlander