Text to Video
#
One API for all video and image models
Always find the best price.
Access Sora, Kling, Veo, Minimax, and 15+ video models through a single interface.
One flat 2% platform fee — vs. 5% on OpenRouter.
The OpenRouter alternative for video & image models
- 2% fee
- flat, all-inclusive
- Full transparency
- into each model's price, latency & quality
Day 0 access to the best model with lowest price. #
Image to Video
Seedance 2.5
Text to Video
Wan 3.0
Image to Video
Kling 2.5
Text to Video
Veo 3
Text to Video
Sora 2
Text to Image
Seedream 5.0 Pro
Text to Image
GPT Image 2
The architecture
Not another router. An optimization layer. #
One endpoint in. Every model, max uptime. We shop Fal, WaveSpeedAI, Atlas Cloud, Replicate, Novita, and more for the best price on every model, every request — and health-check every host continuously so your app fails over automatically and stays online.
Your application
VideoRouter
Inference Optimization Layer
Inside the layer: cost optimization
Six ways we lower your spend. #
We negotiate lower prices
We negotiate volume pricing directly with GPU-hosting providers, so you get rates much cheaper than going direct.
We route to the cheapest host for that model
The same open-weight model — Wan, HunyuanVideo, MiniMax H3 — is hosted by multiple providers at different prices. We route every generation to whichever host is cheapest and healthy right now.
We track live prices, not stale ones
Video/image hosts change per-generation and per-second rates constantly. We resync prices continuously across every provider, so you never overpay against a stale number.
We batch non-urgent renders
Latency-tolerant video and image jobs run through provider batch queues at a discount, and we pass those savings straight through to you.
We serve open-source models ourselves
For select open-weight checkpoints, we run the model ourselves — cutting out the reseller markup entirely.
We charge less
Just a flat 2% platform fee — vs. 5% on OpenRouter — on every render, every model.
The full layer
We hide the complexity. You just get results. #
Everything below runs automatically, on every request, under a single flat 2% fee.
Price, latency, and quality for every provider — in one place, live — so you never have to guess who's cheapest or fastest for the model you want.
Cost optimization
Every render is priced against the cheapest host serving that model at your requested resolution and duration — automatically, not by hand.
Spending control
Set budgets and limits per key, project, or team, and get alerted before spend runs away — not after the invoice.
Model routing
One API for every video & image model — Fal, WaveSpeedAI, Atlas Cloud, Replicate, Novita, and closed hosts like Sora, Veo, and Kling. Swap models with a single string change.
Retry / failover
If a host degrades or rate-limits mid-render, we fail over transparently — your app never sees the outage.
Price, latency & quality, in one place
Every provider hosting a given model, side by side — live price, measured latency, and benchmark quality — so you can see who's actually cheapest and fastest, not guess.
Performance optimization
We track latency and throughput per host and route toward whoever is fastest for your render right now.
Quickstart
Already on the OpenAI SDK? Change one line. #
Point your base URL at VideoRouter and call any open-source or closed video/image model directly — same request format, one API key, a flat 2% fee.
- OpenAI-compatible — no code rewrite
- Works with any host's key, or bring your own
- 2% fee on every request — vs. 5% on OpenRouter
from openai import OpenAI
client = OpenAI(
base_url="https://videorouter.sh/api/v1",
api_key="llmr_sk_...",
)
video = client.videos.create(
model="wan-2.2",
prompt="A drone shot over a neon-lit city at night",
seconds=5,
)
print(video.id)
Plug in once. Let the layer do the rest. #
Create a developer account, grab an API key, and point your OpenAI SDK at us — every video & image host, price, latency, and failover, optimized automatically.
OpenAI-compatible · one API key · 2% fee.