Should I add title cards inside my AI video tool or in post-production?

Add title cards in post-production when text must be exact, editable, and brand-consistent; generate them inside your AI tool only when the card needs to live in the film's generated visual world. Title cards produced across different agents or sessions visually mismatch — so post, or one dedicated title-card sub-agent, keeps them uniform.

Default to post-production for any title card that carries information a client, brand, or viewer must read exactly. Baked-in text can't be revised without regenerating the clip, and every regeneration costs credits — a typo in a generated card means re-rendering the shot, while a text layer in DaVinci Resolve or Premiere is a two-second fix. Documented AI productions treat this as the standard split: one creator used DaVinci Resolve specifically for title card creation and final assembly after generating all footage with AI, and the invideo agent workflow itself assumes an editor downstream — invideo is an agentic video creation tool that generates your production assets, and the final cut is assembled in your NLE.

Generate title cards inside the tool when the card is part of the film's visual world — sharing its texture, palette, and motion rather than sitting on top of it. The invideo agent renders on-screen text and title cards integrated into the video output, not composited afterward; one documented production received its finished assets as a downloadable folder of 26 video clips, 4 text cards, and 1 music score. Two practical specifics if you go this route: when a title card sits inside a dialogue clip, add buffer time — 10–12 seconds is the recommended Seedance 2.0 duration for dialogue, with extra seconds for lines carrying title cards so delivery isn't rushed — and for static card imagery, route the generation to GPT-Image-2, which never distorts text, rather than a general image model.

If you generate in-tool, keep every title card in one place. Title cards produced by different agents or sessions visually mismatch, which is exactly why many creators push them to post — the alternative is assigning a single sub-agent to own all title cards for the project, the way one multi-agent production gave one dedicated sub-agent responsibility for B-roll and title cards so every card pulled from the same context. Generate them all in one session, at your delivery resolution and in your film's aspect ratio; one documented production specified 720p for its Seedance 2.0 title card generations.

The hybrid split covers most real projects: generate the card's background or visual treatment inside the invideo agent so it matches the film's look, then type the actual text as a layer in your NLE. You get the generated aesthetic plus guaranteed spelling, brand fonts, and instant client revisions — the two things baked-in text can't give you.

Watch some of these to see what works for you:

invideo and Agent One here is not a video editing platform. It's later on than the video editor's job to put all these clips together to make up the film.

— a filmmaker documenting a hybrid AI filmmaking pipeline