Run FFmpeg in the cloud
The FFmpeg API for developers and AI agents. Run any FFmpeg command in the cloud, or use a named shortcut like transcode or subtitles. You don't install anything or run a server.
curl -X POST https://api.mediaorca.com/v1/jobs \ -H "Authorization: Bearer sk_live_..." \ -d '{ "operation": "transcode", "input": "uploads/interview.mp4", "output": "public/vertical.mp4", "args": ["-vf", "crop=ih*9/16:ih,scale=1080:1920"] }'What you can build
The media layer under your product
Captions, transcodes, thumbnails, clips, streaming ladders. Every one of them used to mean running your own encoders. Now you ask for the file you want instead.
Add subtitles & captions
Burn captions into the picture or ship them as a sidecar track, so every clip is accessible.
Transcode uploads for the web
Turn whatever users upload into clean, streamable MP4 at the sizes you actually serve.
Generate thumbnails & previews
Pull a poster frame or a scrubbable sprite sheet for any video library, feed, or player.
Convert audio & video formats
Move between MP4, MOV, WebM, MP3, and WAV without thinking about codecs.
Make clips, highlights & GIFs
Trim to a range, stitch moments together, or turn a few seconds into a GIF.
Package video for streaming
Build the HLS ladders and manifests so long videos play smoothly anywhere.
SFX Engine uses MediaOrca to power every sound effect it ships. Each AI-generated master gets trimmed, level-matched, and converted into the formats producers actually work in.
Tens of thousands of jobs later, there's still nothing on their side to scale or keep online.
For agents, and the developers pointing them
Give your agent FFmpeg.
Don't give it a server.
The MCP server hands over every operation as a typed tool, so Claude or Cursor can work with media on its own. Same operations over REST when you'd rather call it yourself.
API & MCP server
One API, two ways in. Write a normal HTTP request, or point Claude, ChatGPT, or your own agent at the MCP server and let it call the operations as typed tools.
Any FFmpeg command
Send the whole command yourself, every flag and filter you need. Or reach for a named shortcut like transcode or subtitles when you just want the common thing.
Zero egress fees
Your files live in storage that doesn't charge you to read them. The egress fee that quietly doubles most media bills isn't part of the deal.
Whole workflows in one call
Chain the steps you need into a single request. Transcode, then thumbnail that result, then package it for streaming, with nothing downloaded and re-uploaded in between.
Webhooks
We post to your endpoint the moment a job finishes, with the signed URL attached. There's no polling loop for you to write or babysit.
Integrations
Wire renders into n8n, Zapier, or Make, or call it straight from your own backend. You send a file and get one back, with no glue code to maintain.
Pricing
You pay for run time and storage
Your monthly price comes back as credit. After that you pay for the seconds your jobs run and the files you keep. Downloading them is free.
Pro
For products and agents doing real work.
Includes $20 of usage
Get started- Compute
- $0.003 / vCPU-min
- Storage
- $0.05 / GB-month
- ✓5 jobs running at once
- ✓1080p inputs up to 60 minutes
- ✓Outputs kept for 30 days
- ✓REST API and MCP server
- ✓Email and community support
Scale
For steady volume, at lower rates.
Includes $99 of usage
Get started- Compute
- $0.0025 / vCPU-min
- Storage
- $0.04 / GB-month
- ✓25 jobs running at once
- ✓1080p inputs up to 120 minutes
- ✓Outputs kept for 90 days
- ✓Priority queue
- ✓Priority support
Enterprise
For large volume and custom terms.
Committed usage, invoiced yearly
Talk to us- Compute
- Volume rate
- Storage
- Volume rate
- ✓Concurrency sized to your load
- ✓Custom limits, 4K on request
- ✓Retention you choose
- ✓SSO, SAML, and an uptime SLA
- ✓Dedicated support and invoicing
Questions
A few things people ask
MediaOrca runs FFmpeg in the cloud so you don't have to. You send a job, either your own FFmpeg command or a named shortcut like transcode, over a normal API, and we run it on managed infrastructure and hand back the finished file. There are no encoders to host and no queue to babysit.