Agent or Timeline? Designing Conversational Video Editing

Should an AI video editing product be a chat Agent or a traditional SaaS tool? In practice, the strongest interface combines both.

Conversation is the intent layer

People naturally describe outcomes: shorten the introduction, keep the product demonstration, remove repeated phrases, add concise captions, or adapt the result for a vertical feed. An Agent can ask for missing context and translate those goals into concrete editing decisions.

But a conversation alone is weak at showing exact state. Users still need to see which footage is ready, what ranges will be kept, how long the result will be, what a render will cost, and whether verification passed.

The workspace is the evidence layer

A purpose-built workspace should keep four things visible:

1. The conversation and tool decisions.
2. The current source or rendered preview.
3. The proposed EDL, cost, job progress, and verification report.
4. A compact timeline that maps every kept range back to its source.

This is not a full manual NLE timeline. It is an inspection surface for an Agent-authored plan.

Rendering must stay deterministic

The model can propose ranges, captions, and animation briefs. A constrained renderer should validate those values, resolve only reviewed presets, keep paths inside the project workspace, and produce the same output when given the same plan. User approval is the boundary between reasoning and execution.

The result is a hybrid product: chat makes complex editing approachable; the SaaS workspace makes it accountable.