Almost all tools for automatic video generation are structured in the same order: first, a script is written, then the text is cut into scenes, and only after that does someone β either a person or an algorithm β assemble the edit from the finished frames. At Telematic, we have consciously reversed this sequence. The timeline of the future video β with scenes, transitions, titles, and a rough storyboard β appears before the final script is generated, and even more so before any frame is drawn.
It sounds like a technical detail, but in practice, it changes the very economics of video production. If the edit is assembled after minutes have already been spent generating images, voiceovers, and subtitles, then any edit β moving a scene, removing a lagging piece, adding a title β means recalculating everything that comes after it. However, if the edit exists as a separate, lightweight layer before heavy generation, the same edits cost almost nothing.
In this article, we will explore what exactly happens at this early stage, why it is not just an engineering convenience but a real tool for controlling the outcome, and how to use it in practice when creating videos through Telematic.
//What is the Editing Phase Before the Script
When the material enters the project β whether itβs an article, a transcription of someone else's video, or a manual text β the system does not immediately write the final text for the voiceover. First, a timeline is formed: a sequence of future scenes with their approximate duration, space for transitions and titles, and a rough storyboard of what will happen on screen. This is a lightweight, quickly generated, and easily editable layer β a kind of skeleton of the video.
On this timeline, you can:
- move scenes around, changing the order of narration;
- cut one scene into two or merge adjacent ones;
- rearrange or remove transitions and titles;
- view a rough run-through in a schematic form, without waiting for images and voice generation.
Only after the structure is fixed does the system unfold a full script beneath it β the text that will go into the voiceover and prompts for the visuals. In other words, the text is adjusted to the approved edit, not the other way around.
//Why This is Not Just About the Order of Operations
The difference becomes noticeable when something needs to be changed. We tested both scenarios on the same materials: in the classic model of "script first, then edit," any shift in structure β say, deciding to move the introduction 20 seconds later β meant recalculating the timings of several scenes' voiceovers, regenerating subtitles for the new duration, and if the project compiles chapters for YouTube, manually recalculating the timestamps.
When the structure is fixed at the editing level in advance, this same edit is simply dragging a block on the timeline. The voice, subtitles, and images for that scene do not need to be touched at all because the scene as a semantic unit has not changed β only its position in the sequence has changed.
There is also a second, less obvious benefit: transitions and titles in Telematic are not an overlay effect but regular scenes that can have their own video and voiceover. Since they exist at the editing level from the very beginning, they can be planned content-wise β for example, placing a meaningful transition exactly where the topic of conversation changes, rather than inserting a title "wherever it fits" after the entire rest of the video has been rendered.
//How This Looks in Real Project Work
In practice, the process for the user looks like this: after the AI has processed the source material into a rough structure, you land on the timeline screen, not immediately in the script text. Here, you can see the number of scenes, their approximate duration, and the placement of pauses. This is the moment to pause and think not about formulations but about dramaturgy: where the video lags, where to shorten, and where, on the contrary, to add a transition to give the viewer time to digest what has been said.
One of our clients, who runs a channel with reviews in a narrow industry niche, shared that he used to spend more time reassembling finished videos than on editing from scratch β because any structural change in his previous tool required regenerating sound and images. After switching to the "edit first" model, such edits now take seconds: moving a block, seeing how it looks on the timeline, and only then launching the generation of heavy content.
//Connection with YouTube Chapters and Narrative Pace
Since the structure of the video is fixed at an early stage, it can naturally be used to build chapters with timestamps for YouTube β a system that viewers see as navigation under the player. Since the scenes and their order are known in advance, rather than being assembled post-factum from a finished video file, the timestamps are accurate and correspond to semantic blocks, not arbitrary cuts.
Similarly, at this level, the narrative pace can be adjusted β dynamic, standard, or calm. This affects how the text will be "laid out" in scenes even before the script is generated: a dynamic pace means more short scenes and transitions, while a calm pace holds longer shots and fewer editing cuts. If this decision were made after the script and voiceover were already ready, changing the pace would mean almost a complete reassembly of the video.
//What This Means for Your Practice
If you regularly work with Telematic, there are several habits to adopt from this practical editing tool:
- Donβt rush to approve the timeline automatically β take a minute to review the order of scenes before launching full generation. This is the cheapest moment for edits throughout the entire production cycle.
- Place transitions consciously, tying them to semantic shifts in the text, rather than leaving the system to decide this later.
- Choose the narrative pace according to the platform's format at the editing stage, rather than trying to "save" an already finished video by shortening lines.
Ultimately, the idea of editing before the script boils down to a simple principle: the structure of the video is a higher-level decision than specific words and frames, and it should be made first. Everything else β text, voice, visuals β is already the realization of this structure, not the other way around. Thatβs why in Telematic, editing does not follow the script but precedes it.