AI Model Wiki

    Skip to main content
    40 min
    Lesson 1 of 5

    Introduction to Text-Based Editing

    From the course Descript: AI Video & Podcast Editing

    Descript fundamentally changes how you edit video and audio by letting you edit media the same way you edit a text document.

    The Text-Based Editing Revolution

    Traditional video editing requires learning complex timelines, keyframes, and audio waveforms. Descript transcribes your media and lets you:

    • Delete words to delete video: Remove a sentence from the transcript, and the corresponding video/audio is removed
    • Rearrange by drag-and-drop: Move paragraphs in the transcript to rearrange your video
    • Type to add content: Insert new narration by typing text (with AI voice)
    • Find and replace: Search for filler words and remove them all at once

    Getting Started

    1. Create a Descript account and download the desktop app
    2. Create a new project and import your video or audio file
    3. Descript automatically transcribes your media (typically 95%+ accuracy)
    4. Review and correct any transcription errors
    5. Start editing by modifying the transcript

    Core Editing Actions

    Removing Content

    • Highlight text in the transcript and press Delete
    • The corresponding audio/video segment is removed
    • Transitions are automatically smoothed

    Rearranging Content

    • Select a paragraph or section in the transcript
    • Drag it to a new position
    • The video/audio follows the text arrangement

    Splitting & Merging

    • Place cursor in the transcript to split at that point
    • Merge sections by removing the gap between them

    Filler Word Removal

    Descript detects and removes filler words automatically:

    • "Um", "uh", "like", "you know", "sort of", "kind of"
    • One-click removal of all filler words
    • Option to shorten pauses and silences
    • Preview before applying changes

    Real-World Example — Ali Abdaal's Podcast Workflow

    Productivity creator Ali Abdaal (3M+ YouTube subscribers) has publicly described using Descript as his primary editing environment for the Deep Dive podcast. His team imports raw interview audio, uses AI transcription to generate the transcript, removes filler words and dead air in bulk, then rearranges interview segments by dragging text blocks. According to his team, Descript cut per-episode editing time from ~12 hours to ~3 hours.

    First Edit Walkthrough — Filler Word Removal

    1. Import video/audio → Descript auto-transcribes (1-5 min per hour of content)
    2. Click Edit menu → "Remove filler words"
    3. Descript highlights every "um," "uh," "like," and long pause
    4. Click "Preview" to hear the result before committing
    5. Adjust: set minimum silence threshold (e.g. remove pauses > 0.8 seconds)
    6. Click "Apply" → all filler words removed with smooth transitions
    7. Timeline auto-adjusts — no manual cutting needed
    

    Total time for a 60-minute podcast episode: approximately 8-12 minutes for the initial cleanup pass.

    Best Practice — Correct Transcription Before Editing

    Descript's editing is anchored to the transcript. If the AI mis-transcribes a word (e.g. "Kubernetes" as "Cube networks"), and you then delete that segment from the transcript, it deletes the wrong audio range. Spend 5 minutes doing a quick transcript review pass before making any cuts — it prevents confusing audio glitches in the final export.

    Common Pitfall — Over-Removing Pauses

    Removing every pause longer than 0.3 seconds creates an unnaturally frantic pace that fatigues listeners. Leave deliberate pauses after key statements — they give listeners time to absorb important points. A good rule of thumb: remove pauses over 1.5 seconds in the middle of sentences, but leave pauses between paragraph-level thoughts.

    Key Takeaways

    • •Text-based editing lets you edit video and audio by modifying a transcript document
    • •Deleting, rearranging, and adding text directly manipulates the underlying media
    • •Automatic transcription with 95%+ accuracy creates the editable text layer
    • •Filler word detection and one-click removal saves hours of manual editing
    0 of 5 lessons marked complete

    Progress is kept in this browser only.