AWS adds TwelveLabs' Marengo Embed 3.0 as a native embedding model in Amazon Bedrock Knowledge Bases, enabling natural-language video/image/audio search
Amazon Bedrock Knowledge Bases now offers TwelveLabs Marengo Embed 3.0 as a selectable embedding model (general availability), letting users create a "Managed Knowledge Base" that automatically ingests video/audio/image files from S3 (and other connectors), generates multimodal embeddings (512-dim vectors) without manual pipeline building, and supports natural-language semantic search/retrieval over that content via console testing or the Bedrock Retrieve API. Available only in us-east-1 and us-west-1; billed at standard Bedrock model invocation rates plus storage/retrieval costs.
Entities: Amazon Web Services, Amazon Bedrock, Amazon Bedrock Knowledge Bases, TwelveLabs, Marengo Embed 3.0, Amazon S3
1 primary
What happened
AWS has made TwelveLabs' Marengo Embed 3.0 generally available as a selectable embedding model inside Amazon Bedrock Knowledge Bases. This lets users build a "Managed Knowledge Base" that automatically ingests video, audio and image files from S3 and other connectors, generates 512-dimensional multimodal embeddings, and supports natural-language search over that content via the console or the Bedrock Retrieve API. It is currently limited to two regions (us-east-1 and us-west-1) and billed at standard Bedrock model invocation rates plus storage and retrieval costs.
Why it matters
This removes a real chunk of engineering work for teams wanting to make video and media libraries searchable by meaning rather than filename or tag: no need to hand-build transcription, frame extraction, embedding and vector-sync pipelines. That is useful for AWS customers already committed to Bedrock, and it strengthens AWS's pitch as the easiest place to build multimodal RAG applications. The impact is moderate rather than transformative: this is a managed integration of an existing third-party model, not a new capability, and the region limits mean many production workloads cannot use it yet.
What is noise
The claimed use cases (sports analytics, security/surveillance, retail, education) are asserted, not demonstrated. No pricing figures, retrieval accuracy benchmarks, or latency numbers are given, so "removes the need to stitch together" is a convenience claim, not a proven performance one. This is standard vendor announcement framing for a routine model-availability update, not evidence of a market shift.
Watch next
- 01Whether AWS expands availability beyond us-east-1/us-west-1 within the next two to three months, signalling real customer demand versus a limited pilot
- 02Published pricing details and any independent benchmarks comparing Marengo 3.0 retrieval quality/cost against a self-built pipeline (e.g. Whisper transcription plus CLIP embeddings)
- 03Adoption signals from AWS re:Invent or customer case studies naming specific production deployments, rather than generic vertical mentions
Coverage
1 storyMore capability signals
Full feed →- Deepseek releases V4.1-Flash, an open-source model that sharply cuts KV cache memory and input-processing compute for AI agents10 Sept 202682
- AI systems outperform expert humans in persuasive communication22 Jun 202681
- WIRED investigation: Flock Safety's AI person-search tools let police run broad description-based surveillance, with weak guardrails against misuse3 Sept 202680
- Google DeepMind launches AlphaGenome Atlas, a free public database of predicted effects for 9 billion possible human genome variants8 Sept 202679