To add subtitles and edit a video with AI, you’ll mainly use two tools together: Descript to transcribe the audio, generate captions, and cut the video by editing text, then Runway to generate any shots you’re missing. What you get is a captioned rough cut that you still review pass by pass before delivery.
Transcribe the video and generate captions
Start by uploading your video to Descript and letting it transcribe the audio, then generate captions from the transcript in one step. Descript is a practical choice for transcript-first podcast and video editing; it uses a freemium model with a free plan, and paid tiers are listed on its official site (per the tool’s official site, verified 2026-07-15).
Transcription can miss names and specialist terms, and people’s names, brand names, and jargon are the usual culprits. Read the transcript through and fix it before you cut anything.
Cut the video by editing the transcript
In Descript, delete a word in the transcript and the matching footage drops out with it, so you can remove filler, verbal tics, and repeated passages by editing text. You locate cuts by reading the script instead of scrubbing the timeline.
Deleting text removes the audio, but the picture and tone don’t always join up. Re-watch each cut point for jumps in mouth movement, gesture, or background sound.
Generate the shots you’re missing with Runway
When you’re short a cutaway or a transition, use Runway to generate concept footage from a text description and drop it into your timeline. It’s useful for concept footage and iterative shot development, and it uses a freemium model with a free plan, with paid tiers listed on its official site (per the tool’s official site, verified 2026-07-15); which models you get and your export entitlements vary by plan.
An empty city street at early morning, soft natural light, slow panning camera, photorealistic
Generated motion can lose continuity, so object shapes and lighting direction may not match shot to shot, which shows when a clip sits next to real footage. Judge each clip in context, played alongside the real shots around it, not on its own.
Check before you publish
Before you export, run through the captions, your cut transitions, and the rights on your footage, since none of these are things the AI can sign off for you.
- Read the captions once more and confirm names, brands, and terms are corrected, because Descript’s transcription gets these wrong
- Re-watch every cut point so the picture and tone don’t jump
- View each Runway-generated shot in context, next to the real footage, for continuity
- Clear rights separately for people, brands, and any supplied media, and note that your media and AI access vary by plan