How do I edit videos with text prompts, just by describing what I want?

To edit videos with text prompts, open an editor that has an AI assistant, add your footage, and type what you want in plain English, like "add subtitles" or "edit this down to 2 minutes." The AI makes the edits. You review the result, then refine it with another prompt or fix small details by hand on the timeline.

Key takeaways

• Prompt-based editing means you describe the edit and the AI does the clicking. Common requests include captions, resizing for TikTok, adding music, removing silences and cutting a long video shorter.
• Good prompts give context: what the video is for, where it will be posted, how long it should be, the tone, and who's watching.
• The AI can get things wrong. Descript, Kapwing and CapCut all say you should review what it makes.
• A timeline editor is still the better choice when you need precise control over timing, layers or effects.
• ZippyClip lets you edit in plain English or by voice, shows the AI's reasoning as it edits, and can reuse a saved prompt on every video you drop into a watch folder.

What does "editing with text prompts" actually mean?

You give the editor instructions in plain language and it does the edit for you. There's no hunting through menus.

In most tools, the assistant sits in a chat panel next to your project. VEED's AI Agent is in the top-right corner of a project. Descript's Underlord opens from the right-hand sidebar. Kapwing's assistant, Kai, is a chat where you can attach video, images, audio or documents.

Don't confuse this with text-based editing. That's where you edit a written transcript and the video gets cut to match, a feature found in tools like Adobe Premiere and Descript. Prompt editing goes further: you describe the result you want, not each cut.

What can you ask an AI video editor to do?

You can ask for most of the repetitive work in editing. Real example prompts from the tools' own pages:

• "Add subtitles to this video" (VEED)
• "Adjust the aspect ratio for TikTok" (VEED)
• "Make this video work for Instagram." (VEED)
• "Hide my jumpcuts by adding zooms" (Descript)
• "Edit this down to 2 minutes—focusing on when I talked about Freud" (Descript)
• "Make the trailer for this webinar recording" (Descript)
• "Copy all the parts where Laura is talking into a new composition" (Descript)

Beyond those, the tools list jobs like these:

• Captions: social-style captions with the spoken word highlighted in your brand colour (Descript).
• Clips: split a video into parts under a set length, or find short social clips and make them vertical (Descript, Kapwing).
• Audio: remove silences (VEED), clean up sound or lower background audio so speech stands out (Descript).
• Translation: translate subtitles or the whole video (Descript, Kapwing).

For more on these two jobs, see our guides on adding captions automatically and turning long videos into short clips.

How do you edit videos with text prompts, step by step?

The steps are much the same in every tool:

1. Pick an editor with an AI assistant. Options include ZippyClip, Descript, Kapwing, VEED and CapCut.
2. Add your footage and open the assistant's chat panel.
3. Describe the edit with context: what the video is for, where it will go, how long it should be, the tone and who it's for.
4. Let the AI work, then watch the result.
5. Refine it. Send a more specific follow-up ("only the part about pricing"), or open the timeline for small fixes.
6. Save prompts that work so you can reuse them next time.
7. Export, and look out for watermarks on free plans.

How do you write a good editing prompt?

Give the AI context, not clever wording. In a guide updated in July 2026, Descript says results depend more on how much the editor knows about your goal than on how you phrase the prompt. If the result misses, add detail before you start rewording.

Descript's own before-and-after shows the difference.

Weak prompt:

"Cut this into a short video."

Strong prompt:

"Turn this into a 45-second TikTok with an upbeat tone for an audience of video game fans. Include captions, a fast cut pace, and a punchline at the end. The viewer should laugh and decide to follow my account for more."

A simple checklist for every prompt:

• Purpose: a teaser, a product announcement, a tip video?
• Platform: TikTok, Reels, Shorts, YouTube?
• Length: "under 60 seconds," "2 minutes."
• Tone: upbeat, calm, dramatic.
• Audience: who's watching.
• Success: what you want viewers to do (laugh, click, follow).

Pointing at specific scenes ("the part where I demo the app") also gives the AI better context, according to Descript.

What are the limits of prompt-based editing?

The AI can misunderstand you, so you always need to check its work. Descript says Underlord "can make mistakes, overpromise, or make incorrect assumptions." Kapwing says generative AI "is rarely 100% correct." CapCut calls its chat results starting points that need review.

Other limits to know about:

• Credits cost money. Descript and Kapwing charge AI credits when you use their assistants. Descript says the cost of the same prompt can vary between runs. Kapwing says generations you don't like still use credits.
• Not everything is available by prompt. Kapwing notes that some tasks can't be done through Kai and have to be done in its editor.
• Some inputs aren't supported. Descript's Underlord can't open a link you give it, so you paste the text in instead.
• Setup can vary. CapCut's prompt editing runs through a plugin for the ChatGPT desktop app on PC, and availability depends on your app version, account and region.

When is a timeline editor still better?

Use a timeline when you need exact control. A guide from Reduct (a company that sells text-based editing) says timeline editing is better for fine visual control, effects, colour grading, transitions, layered adjustments and footage with little speech. Text-based editing is better for talk-heavy recordings, removing filler words, and beginners.

That's why it helps to have both. Kapwing opens Kai's result in its Studio for timing, layers and text, and CapCut lets you move from the chat to its full editor. Let the prompt do the heavy lifting, then nudge the last details by hand.

Which tools let you edit videos with text prompts?

ZippyClip is made by the team behind this blog. Prices below are as of October 2026 and can change.

Tool	Best for	Key prompt-editing features	Price / free option	Watch-outs
ZippyClip (our pick)	Businesses and creators who want edits done and posted for them	Plain-English or voice prompts, visible AI reasoning, saved prompts for watch folders, real timeline	Free download with 50 API tokens (watermark); Pro removes the watermark	Desktop app (Windows and macOS); free exports carry a watermark
Descript (Underlord)	Talking-head videos, podcasts, webinars	Chat co-editor, bulk edits, saved skills, transcript editing	Free $0 (720p, watermark, limited Underlord); Hobbyist from $16, Creator from $24 per person/month	Beta; credit cost varies per prompt
Kapwing (Kai)	Browser editing mixed with AI-generated media	Edit or generate by chat; result opens as a layered project	Free $0 (10 credits, 720p, watermark); Pro $16, Business $50	Tight free plan; not every tool works through Kai
VEED (AI Agent)	Quick social edits in the browser	Subtitles, resizing, music, silence removal by prompt	Free $0 (watermark, 50 AI credits/month); Creator $25/month or $12.25/month billed yearly	Plans page doesn't say which plan includes the agent
CapCut (via ChatGPT desktop)	People who already use ChatGPT and CapCut	Chat drafts and refinements, then CapCut's full editor	Free $0; Pro $19.99/month	Needs a plugin; availability varies by region and account

1. ZippyClip: our pick for editing by describing what you want

ZippyClip is our pick because it goes beyond doing the edit. It also keeps editing and posting for you once you know what you want.

• Edit in plain English. For example: "Cut this long video into 3 viral shorts" or "Remove the dead air and schedule for tomorrow at 9am."
• Dictate edits by voice when typing is slower.
• Watch the AI's reasoning as it edits, so you can see what it's doing and why.
• Save a prompt for a watch folder. Point ZippyClip at a folder on your computer. Every video you drop in gets edited with your saved prompt and published to the channels you picked. You can also send a clip to Google Drive from your phone, and your desktop picks it up.
• A real timeline (drag, trim, layer) for when you want manual control.
• The edits people ask for most: word-level captions, smart vertical reframing, pause and filler-word removal, and viral-moment detection for long recordings.
• Publishing built in to YouTube, TikTok, Facebook, Instagram and X, with AI-written titles, descriptions and hashtags. TikTok publishing is a Pro feature.

It's a desktop app for Windows and macOS. Your videos and projects stay on your computer. It's free to download with 50 API tokens, and free exports include a watermark. Pro removes the watermark. If you want the whole process hands-off, see how to put your video marketing on autopilot.

2. Descript (Underlord)

Underlord is a chat co-editor inside Descript, and it's still labelled beta. It handles bulk jobs like captions, zooms, translation, trailers and pulling out one speaker's parts. You can save repeated instructions as skills. Paid plans unlock premium AI models.

Watch-outs: each prompt uses AI credits, and the cost can vary between runs. The Free plan exports at 720p with a watermark and only limited Underlord access. Descript's pricing page shows two prices per plan without labels and says annual billing saves up to 35%.

3. Kapwing (Kai)

Kai is Kapwing's chat assistant. It can edit footage you upload, for example adding translated subtitles, finding highlights and resizing them for TikTok, or merging clips. It can also generate new media. The result opens as an editable project in Kapwing Studio, and Kapwing shows an estimated credit cost before you generate.

Watch-outs: the free plan includes 10 credits, a watermark, 720p exports and a 4-minute limit per export. Not every tool can be run through Kai.

4. VEED (AI Agent)

VEED's AI Agent handles everyday social edits from short prompts: subtitles, aspect ratio, music and silence removal. It can also generate clips with models such as Veo 3.

Watch-outs: VEED's AI agent page says the AI playground needs "a subscription," and its plans page doesn't say which plan includes the agent. The free plan is watermarked.

5. CapCut (with the ChatGPT desktop app)

CapCut's plugin for the ChatGPT desktop app on PC lets you describe a video or upload a draft. You can then refine it in chat (shorten scenes, reorder clips, adjust text timing, change music) and finish in CapCut's editor.

Watch-outs: you have to install and authorise a plugin, and availability varies by app version, account and region. CapCut calls the results starting points.

Is editing videos with text prompts free?

Usually you can try it for free, but serious use costs money. Kapwing, VEED, Descript and CapCut all have $0 plans. Their free tiers come with watermarks, small AI-credit allowances or limited assistant access. ZippyClip is free to download with 50 API tokens, and Pro removes the watermark.

Can you edit videos by voice?

Yes, in ZippyClip you can dictate your edits by voice instead of typing them, then watch the AI's reasoning as it works. We didn't find voice prompting on the official pages we checked for the other tools.

Sources

• Descript Help: Underlord (beta), your AI co-editor
• Descript: Your all-in-one video agent
• Descript: Pricing
• Descript blog: The secret to better AI video editing: context, not prompts
• Kapwing: Pricing
• Kapwing: FAQ
• Kapwing blog: Giving AI a canvas to paint on
• VEED blog: Introducing AI Agent by VEED
• VEED: AI Video Agent
• VEED: Plans and prices
• CapCut × Codex
• CapCut: Pricing
• Adobe Help: Overview of text-based editing
• Reduct: Timeline vs text-based video editing