Agent Chat

The Flova Agent is the operator inside your creative project. You can tell it what you want in natural language from the chat box on the right. It reads the current project state and decides what to do next.

It is not just a chat assistant. It is a system that coordinates different creation modules to execute production tasks for you.

What the Agent can do

• Discuss any question with you during the project creation process
• Design scripts, shot lists, or reference media, then create the storyboard
• Analyze the images, videos, audio, or documents you specify
• Generate image, video, music, and voiceover media
• Manage the relationship between media and the storyboard, and bind media to elements, shots, or audio layers in the storyboard
• Assemble the editing timeline from the storyboard and existing media

How the Agent works

The Agent automatically coordinates different modules based on the task type. You usually do not need to name these modules. Just describe the goal and constraints.

• Planning: understands your goal, judges task complexity, decides whether to create a Final Video Spec or load a Skill, and arranges the next steps.
• Media understanding and processing: analyzes uploaded videos, images, music, or documents. When needed, it can also extract video clips, key frames, or audio as references for later storyboard and generation work.
• Storyboard design: creates or edits key elements, shots, and audio layers, forming the structural foundation for later media generation and editing.
• Media generation: generates or regenerates images, videos, music, and voiceovers, while managing assets, versions, and storyboard bindings.
• Editing assembly: creates or modifies the editing timeline based on the storyboard and selected media, handling clip order, rhythm, volume, and audio-visual sync.

How to ask the Agent

The clearer your request, the easier it is for the Agent to produce results that match your expectations. You can add information from the following angles.

Say what you want to make

Turn the story of Cinderella into an AI short film of about 3 minutes, horizontal 16:9, English dialogue, 3D animation style.

Add key specifications

Create a 30-second vertical earbud ad that highlights noise cancellation and lightness, with no dialogue and fast-paced BGM.

Explain how media should be used

Refer to the product images and BGM I uploaded. First, design several ad concepts for me. Do not generate anything directly yet.

State constraints

Do not rewrite the script I uploaded. Organize the storyboard directly from the original text.

Clarify whether you want discussion or execution

Do not generate media for this project yet. First, tell me what steps you plan to take.

Agent, Skill, and Final Video Spec

The Agent reads the project context and decides how to move forward.

• Final Video Spec: records global specifications for the current video, such as aspect ratio, language, duration, and style
• Skill: defines how this type of video should be made, such as whether to design the storyboard first, how to use references, which models to use, and what prompt rules to follow
• Agent: reads this context and coordinates subsequent storyboard, media generation, and editing tasks

If you are only generating an image, a video, or a single media task, the Agent usually does not need to create a full Final Video Spec or load a Skill. If you clearly want to make a complete video, the Agent usually creates or loads this context first so later steps remain consistent.

Is the execution process visible?

When the Agent is working, you can see what step it is advancing in the conversation, for example:

• Analyzing media
• Creating the storyboard
• Generating key elements
• Generating shot videos
• Assembling the editing timeline

At the same time, results appear in the corresponding panels:

• Storyboard results go into the storyboard area
• Image, video, and audio results go into the Media Files Panel and the media area under the storyboard
• Final editing results go into the timeline
• Document updates can be viewed in the Docs panel

This lets you watch the Agent's execution process while checking results in the left panels.

Stop, revert, and branch

• If the direction is wrong during execution, stop the current task, add more information, and continue.
• At the bottom of each AI message, choose back to this moment to restore the project to the state of that message. You can also choose Brunch in new project to create a new creative direction from a specific moment and compare different versions.
• This is useful when story direction, shot style, or media selection diverges.

Work with manual editing

The Agent and manual editing can be used together.

• The Agent is better for multi-step, cross-module tasks that require planning.
• Manual editing is better for small adjustments where you know exactly what to change, or for media management work.
• You can let the Agent continue generating the next batch of content while organizing existing media, editing the storyboard, or using comment generation on the left.
• If parallel changes affect the same content, the system opens a conflict panel and lets you choose which version to keep.

Usage tips

• For complex projects, let the Agent plan first, then enter batch generation.
• Before media generation, check whether the Final Video Spec and Skill match your requirements. These two are the foundation for complex video tasks.
• When uploading media, explain their purpose clearly, such as character reference, scene reference, product image, or music rhythm reference.
• State directly what should not be changed, such as "do not rewrite the script" or "do not redesign the character".
• Check important milestones in time, such as the storyboard and key element images.
• Small edits can be done manually instead of sending everything to the Agent.