Here's the breakdown of how Fliki stitches together your text into a finished video. Think of it as a three-stage pipeline: ingestion, assembly, and rendering.
**Stage 1: Text Processing & Intent Mapping**
You paste your script. Fliki's NLP engine first parses it, looking for keywords, named entities (like "Eiffel Tower"), and sentiment cues. It uses this to decide on two core things:
* **Scene Selection:** It matches keywords to its media library (stock images/video clips). The word "sunset" triggers a search for relevant sunset footage.
* **Voice Synthesis:** Your text is chunked into logical phrases and sent to a Text-to-Speech (TTS) engine. You choose the voice, but Fliki handles the timing and pronunciation.
**Stage 2: Automated Assembly**
This is where the "magic" happens, and it's essentially a templating engine. Fliki creates a timeline based on simple rules:
* Each sentence or paragraph typically gets its own scene.
* The duration of the scene is dictated by the length of the generated audio clip for that text segment.
* The selected media (image or clip) is placed underneath that audio, often with basic Ken Burns effect (slow zoom or pan) applied to prevent a static look.
* On-screen text (subtitles or highlights) is auto-generated and synced to the audio.
**Stage 3: Rendering & Delivery**
The platform takes the assembled timeline—audio tracks, visual assets, text layers, transitions—and uses its rendering farm to encode it into a single video file (like an MP4). You then download or share it directly.
The critical thing to understand is that you're not getting a custom AI that *understands* your narrative. You're getting a system that uses keyword matching against a media database, couples it with TTS, and applies a consistent assembly template. For complex narratives, this often leads to generic or oddly matched visuals unless you manually intervene and select each piece of media yourself.
plan before apply