MGT · Multimodal Generative Technology.
The intelligence layer under everything Aive does: MGT combines 25 models to link what is seen, what is said and what it means, tied to the exact second.
From source video to the next decision.
A 38-second storyboard in nine steps. Pause, step through or read the transcript.
A source asset, in its context.
A video master comes in with title, language, owner, date and rights.
Rules come before scale.
The Playbook overlays format, market, brand rule and approver.
MGT structures the master.
MGT turns the master into cards: versions, formats, available scores and destinations.
Proof stays linked to the source.
Zoom on a moment and its timecode when available.
One idea, many contexts.
Three versions are built for three channels.
AI prepares. Your team decides.
A person approves a version and its destination.
Read the context, not just a score.
An observed AI answer opens with URL or domain, question, date and coverage.
Every result keeps its level of proof.
Social, GEO, Referral, Engagement and Conversion rails stay separate.
Your content speaks for you. You stay in control.
Back to the Playbook and the decision log.
Animation transcript
- 0–3 s · Master. A video master comes in with title, language, owner, date and rights. A source asset, in its context. (An identified asset, with its rights.)
- 3–7 s · Playbook. The Playbook overlays format, market, brand rule and approver. Rules come before scale. (Governance before generation.)
- 7–12 s · MGT. MGT turns the master into cards: versions, formats, available scores and destinations. MGT structures the master. (Scoring, versioning, distribution and GEO.)
- 12–16 s · Moment. Zoom on a moment and its timecode when available. Proof stays linked to the source. (Missing or unanalysed modalities are clearly flagged.)
- 16–21 s · Versions. Three versions are built for three channels. One idea, many contexts. (Every version carries its status.)
- 21–25 s · Approval. A person approves a version and its destination. AI prepares. Your team decides. (No implicit automatic publishing.)
- 25–31 s · AI answer. An observed AI answer opens with URL or domain, question, date and coverage. Read the context, not just a score. (A protocol signal, not a visibility guarantee.)
- 31–35 s · Distribution. Social, GEO, Referral, Engagement and Conversion rails stay separate. Every result keeps its level of proof. (Citation ≠ click; referral ≠ causality.)
- 35–38 s · Log. Back to the Playbook and the decision log. Your content speaks for you. You stay in control. (A governed, reviewable loop.)
What is MGT?
MGT is Aive’s proprietary multimodal technology. It structures a video master for Aive’s published uses: scoring, versioning and distribution. The outputs, modalities and performance shown always match the enabled module, the analysed corpus and verifiable product proof. No arrow in the chain bypasses human decision.
- Master
- The source video, with its context, owner and rights.
- Playbook
- Brand rules, formats, markets, rights and approvals applied before any generation.
- Version
- An adaptation of one idea for one channel, carrying its status.
It all starts with the data inside your videos.
MGT, our proprietary multimodal technology, turns every video into structured data. The same data powers social performance in Aive Studio and AI visibility in Aive GEO.
- Campaign masters and ads
- Rushes, episodes and long-form
- Demos, interviews, podcasts
- Archives and back-catalogue
- Scenes, shots and pacing
- Objects, products and logos
- Voice, transcript and emotion
- Speakers, sourced statements, timestamped moments and entities
- Studio: Creative Score and per-platform compatibility
- Studio: versions, cut-downs, captions, distribution
- GEO: citable articles, pages and structured data
- GEO: citations, sources and sentiment measured
Most AI skims video. MGT reads it.
Scenes, faces, objects, products, on-screen text, speech, speakers, sourced statements, meaning: MGT brings it all into one structured representation.
General-purpose AI gets a rough summary; MGT gets an operable understanding: versionable, searchable to the second, scorable, citable.
LLMs are getting better at watching video. MGT was built to operate on it.
Frontier technology. Enterprise grade.
Seven years of multimodal video AI built in-house: MGT combines 25 models to read every frame. Built with NVIDIA. Published SOC 2 Type II report, GDPR compliant, client data never used to train models.
Eight layers, one system.
Playbook
Goals, brand rules, markets, formats, rights and approval.
MGT
25 models structuring the master, to the exact second.
Aive Studio
Creative Score, versioning, distribution and Storyboard, from the master.
Video GEO
Timestamped moments, sourced statements, pages, structured data and Content Graph.
Index and Aive Expert
Corpus, index, search and sourced answers, pointing back to the video moment.
Distribution and measurement
Social, CMS, Player, media, AI answers and first-party data.
Controls
Rights, roles, review, logs, cost, publishing, rollback and limits.
Security
SOC 2 Type II, GDPR, encryption, no cross-customer training and ISO 27001 in progress.
A public matrix, method included.
Measured: dated observation protocol. Monitored: regular observation, partial coverage. An API answer is never presented as the experience of every user.
| Environment | Status | Method | Proof required |
|---|---|---|---|
| ChatGPT | Measured | Dated observation protocol | Model, date, parameters, region, cadence |
| Gemini | Measured | Dated observation protocol | Same level of proof |
| Claude | Measured | Dated observation protocol | Same level of proof |
| Perplexity | Measured | Dated observation protocol | Same level of proof |
| Google AI Overviews | Monitored | Observation, partial coverage | Country, interface, query, date, method |
Same matrix as the Aive GEO hub; method detailed in the Methodology.
NVIDIA, at the published level of proof.
Aive announced its presence at NVIDIA GTC Berlin 2026. In an official post dated 16 September 2026, Aive states that it launched a Video GEO solution with NVIDIA.
Any broader presentation of a technology collaboration will be published with a joint source, an explicit scope and the corresponding trademark rights.
Enterprise-grade security and compliance
Secure by design: customer data is never used to train models, and intellectual property remains fully with the customer. SOC 2 Type II · GDPR · encryption in transit and at rest · access controls · logs. ISO 27001 in progress.
Is MGT a generative AI model?
Aive presents MGT as a proprietary multimodal technology that structures video into usable data for scoring, versioning and distribution. Published capabilities match documented modules.
Which modalities are analysed?
Modalities depend on the analyses enabled for your workspace: image, sound, speech, text and metadata. An unanalysed modality is clearly flagged.
Who approves generated content?
Your team. AI prepares versions and recommendations; publishing stays subject to your approval.
Where is data processed?
EU/EEA processing is prioritised for core services; governed transfers are possible depending on policy and contract.