ElevenLabs just turned video production into a conversation

ElevenLabs launched Flows Agent, a conversational AI assistant embedded inside the company's node-based creative workspace, ElevenCreative. The agent takes a natural language description of what a user wants to create, a podcast intro, a product ad with a lip-synced avatar, a narrative video with layered sound effects, and builds the entire pipeline from scratch, selecting models, connecting nodes, and executing the generations.

The move turns what has been a drag-and-drop visual programming interface into something closer to a collaborative conversation. And it fits a broader pattern across the generative AI landscape: the hard part is no longer generating a single piece of media, but chaining multiple generations together into something coherent. See also: how AI agents hit walls when sequencing creative steps.

From visual programming to chat interface

ElevenCreative Flows was already a capable node editor that connected over 50 image and video models alongside ElevenLabs' own audio models into one workspace. Users could generate an image, animate it, add a voiceover, lip-sync it, and score it with music and sound effects, each step feeding into the next. The interface was visual, but it still required understanding the logic of nodes and connections.

Flows Agent replaces that learning curve with a chat side panel. Describe what you want, and the agent selects the models, creates the nodes, connects them, and runs the workflow. Before executing expensive operations, it asks clarifying questions, confirming video length, tone, structure, to reduce wasted credits. This mirrors the trend spotted in Google Vids turning a selfie into a full production studio.

The agent also handles iteration through conversation. A user can say "try a warmer voice", "swap the background to an office", "make a version in Spanish", or "use Veo instead of Kling", and the agent modifies the existing pipeline and re-runs it. The cost of exploring a new creative direction drops to a single sentence.

Assist mode and background execution

An assist mode gives users approval control over costly operations. When enabled, the agent pauses before running expensive generations and asks for confirmation. The user stays in control of credit consumption while the agent handles workflow construction and orchestration. Background execution means users can start a complex workflow, close the tab, and come back to a completed output. The agent also works alongside multiplayer features: collaborators in the same workspace see the agent building nodes and running generations in real time.

The bigger picture: agentic orchestration

Flows Agent is a concrete example of a shift visible across the generative AI industry for months: multi-modal generation has become routine, agentic orchestration is replacing manual chaining, and real-time feedback is turning generation from a gamble into a conversation. ElevenLabs is betting that the bottleneck in video production is no longer the quality of any single model but the friction of assembling models into a coherent pipeline. If that bet is right, the next competitive moat in generative media won't be a better text-to-video model, it will be an agent that knows how to combine a dozen mediocre ones faster than a human can wire up a single workflow manually. For a deeper look at this architecture, see the 96% context cut that finally makes AI agent tool hierarchies work.

What it means for creators and marketers

For individual creators and marketing teams, Flows Agent lowers the technical bar to multi-modal production. The tool is available today inside ElevenCreative at elevenlabs.io/app/flows. The real open question is whether an agent that abstracts away pipeline construction also abstracts away the creative insight that happens when you manually tinker with each node. Some of the best outputs come from unexpected combinations that a linear agent, optimized for plausible pipelines, might never suggest. ElevenLabs may need to balance ease of use with supporting open-ended creative exploration over time, as explored in the practical guide to effective prompting.