AI media intelligence
AI that understands your video. And puts it to work.
More than 30 AI models watch and listen to every hour of your content so your teams can find it, reuse it and turn it into new stories. For broadcast, media and entertainment, in the cloud or on your own servers.
In production
RTVE, the Congress of Peru and Gestmusic rely on Amplify.
Public broadcasters, parliaments, production companies and institutions already working on their media with our technology.
The problem
You record more than anyone can watch.
The archive grows every day and describing it by hand does not scale. What is not described cannot be found. And what cannot be found is never used again.
38 % of the archive described
The journey of your content
From archive to air, inside your own systems.
Follow an hour of video from the moment it arrives until it airs again. At every stage, Amplify does the heavy lifting and your team decides.
01IngestPlatform
Bring it in from anywhere.
Single files, decades of archive or a live signal. Amplify takes it from your storage, over the API or by streaming, without making you move anything.
- Local, Samba, FTP, S3, HTTP and LTO tape
- Live signal over SRT and RTMP
- REST API, chunked upload and desktop app
02UnderstandPlatform
It watches and listens to everything.
Every asset comes out described: who speaks and what they say, who appears, what is written on screen, which brands, objects and scenes are there, and what it is about. All of it timecoded.
- Transcription, speakers, language and translation
- Faces, on-screen text, logos, objects and scenes
- Entities, summary, classification and topic segmentation
The Amplify platform analysing a news programme: transcript, entities, faces and objects 03FindSeiri
Ask. Or hand it to the agent.
Search the way you talk and jump to the exact second. And when the task is long, agent mode plans it, shows you every step and asks for confirmation before touching anything.
- Natural-language search across the whole archive
- The agent’s plan visible, step by step
- Human confirmation before every action
04AutomatePlatform
Write the rule once. It runs every time.
You write in plain language what should happen and the platform turns it into a workflow with a trigger, conditions and actions: a complete process, with no programming. You review and refine it in a visual editor.
- Workflows: triggered when a video is processed or edited, with AI-evaluated conditions
- REST API: your applications ask the platform for what they need, when they need it
- Plugins: run an action when a system event occurs
- Webhooks: the platform notifies your systems as soon as a process finishes
05CreateNews
From search to a finished cut.
News writes the script, picks the shots and soundbites from your archive, voices the narration and delivers the sequence in your editor. The journalist reviews every decision before export.
- Block-by-block script in your editorial line
- Real shots from your archive, timecoded
- EDL and XML export to your editor
News building a story from the archive 06BroadcastSeiriVoice Live
Live, and in several languages.
SeiriVoice Live transcribes, translates, captions and dubs the signal as it happens. And everything above flows back into your systems: the MAM, the editor, the OTT.
- Subtitles, closed captions and prompter view
- Simultaneous translation and synthetic dubbing
- Input over SRT or RTMP
The engine
One platform under everything: analyse once, use everywhere.
The Media Understanding Platform (MUP) is the platform everything else runs on. It analyses each file once and puts the result within reach of an agent, your workflows, your plugins and your MAM.
Agent, workflows, plugins and API: how it is builtListens
Voice
- Transcription in 90+ languages
- Speakers and voices
- Translation and subtitles
- Entities and topic segmentation
Watches
Vision
- Faces, logos and objects
- On-screen text
- Shots and scenes
- Visual tags
Thinks
Mind
- Audio and image together
- Semantic search and chat
- Agent mode
- Workflows in plain language
- AI models working together
- 30+
- languages transcribed
- 90+
- integrations with MAM, editing and storage
- 15+
Products
Three products on the same platform.
Archive, live and newsroom each work in their own tool, and all three build on what the platform has already understood.

Understands and finds
Seiri
Natural-language search, timecoded metadata and cataloguing for your archive, inside your asset manager.
For archive and documentation
Meet Seiri
Live, in more languages
SeiriVoice Live
Transcribes, translates, captions and dubs the signal as it happens.
For broadcasts, sessions and events
Meet SeiriVoice Live
Creates stories
News
From footage already understood to a story with script, shots and voice-over, ready for your editor. The journalist stays in control.
For newsrooms
Meet NewsFor productionsReality, entertainment or documentaries? KAI brings this intelligence to workflows built for the pace of a production. Meet KAI
Results
In production, with numbers.
Live projects, every figure with its source.

Archive · public broadcasterRTVE
28,000h/year28,000 hours of archive a year: 18,000 of Spanish audio, 5,000 of image content and 5,000 multilingual, with Spain’s co-official languages translated into Spanish. Service delivered by Crosspoint with Seiri at its core.
Read the project
Parliament · with MediaflexCongress of Peru
160TB160 TB of historical media and the channel’s daily production in a single environment. Seiri’s metadata is generated at ingest and used directly inside the MAM. Integrated by Video Broadcast; fully operational.
Read the project“We chose TMD’s Mediaflex and Amplify’s Seiri because each is a strong solution in its own right, but especially because of what they can achieve together.”
“The integration with Seiri takes that further by making AI-generated metadata immediately available and useful within the MAM.”
Why Amplify
Built with media teams, for media teams.
Your content stays put. So does your MAM.
- 01
Cloud, hybrid or your own servers
As a cloud service, hybrid, or on your own air-gapped servers, with models running locally. You decide where your content lives.
- 02
Everything is yours
The metadata, the embeddings and the face, voice and logo models you train. Export them whenever you want.
- 03
Designed for compliance
EU AI Act, GDPR, accessibility and ENS by design. Two-factor login, AD/LDAP and an audit trail for every action.
- 04
Your team has the final say
The AI proposes and your team validates. Every field, every shot and every agent action can be reviewed before it is applied.
- 05
Inside your MAM
Documented REST API and Developer Portal. Metadata lands in your asset manager: no extra window to learn.
- 06
Predictable pricing
Perpetual licence or subscription, with no surprises per processed minute. Entry pricing is public.
See pricing
Who it is for
Every team asks it for something different.
Archive and documentation
Decades of holdings described without watching them. Your team validates; it does not type.
SeiriNewsroom
From a search to a story with script, shots and voice-over, ready for your editor.
NewsParliaments and institutions
Sessions and events searchable by speaker and topic, and captioned live.
SeiriVoice LiveProduction companies
A production’s rushes, searchable in the team’s own vocabulary.
KAIIntegrators and MAM vendors
An API so your platform can ask what its users ask.
Platform
What we get asked
Questions from the person who signs it off.
Does it work with my languages and my voices?
At RTVE, Seiri transcribes Catalan, Galician, Basque, Valencian and Balearic and translates them into Spanish. We measure accuracy on your voices with you, on a sample of your archive, before you buy anything.
Does my content leave my building?
Only if you want it to. Amplify installs on your servers with local models, and the on-premise licence verifies offline. Cloud service and hybrid deployment are available too.
How much work is integrating it with my MAM?
With TMD’s Mediaflex the integration is already in production. With other systems, a documented REST API and a Developer Portal with a test console.
What if the AI gets it wrong?
Every piece of data carries its timecode: it is checked in one click and corrected before it reaches the MAM. The agent shows its plan and asks for confirmation before acting.
What happens to my archive team?
The AI removes the viewing and the typing. The judgement stays with your archivists, who validate, correct and train the models on their own taxonomy.
Am I locked in?
Your metadata leaves whenever you want as CSV, Excel, JSON or over the API, and the AI models are interchangeable.
How to start
Try it on your own video.
No generic deck: we look at it working on your footage.
- 01
You write to us
Someone from the product team replies within one or two working days.
- 02
30 minutes on your workflow
A session focused on your archive, your MAM and what you need to find.
- 03
Your sample, described
If it fits, we analyse a sample of your content and measure the result with you.

Next step
Show us your content. We’ll show you what it’s worth.
A 30-minute demo on your own footage, with someone on our team who has worked in television.

















