← Home
PROMPT GUIDE

By Valmera Editorial · Published · Updated

Edit Video by Chatting With AI

You can edit a video by describing the changes in plain English. In Valmera, upload your footage and type requests like “cut the ums and add captions.” An AI agent performs them on your real footage, renders a preview, and you keep chatting to refine it. No timeline is required, though one is there when you want it.

This guide covers how natural-language editing works, a prompt formula, 34 example prompts mapped to the tools the agent calls, and how to phrase revisions. The product itself is on the Valmera AI video editor page.

How to Edit a Video by Chatting With AI

  1. 1
    Upload your footage
    Add a video up to 100 GB and 3 hours (MP4, MOV, M4V, MKV or WebM). Valmera indexes it once: a word-level transcript with speaker labels, shots, silences and frames.
  2. 2
    Type what you want
    Describe the change in plain English, one request or several at once: “Cut the pauses, add captions and make it 9:16.”
  3. 3
    The agent edits and previews
    It plans the steps, calls the editing tools, renders a preview and checks it before replying with what it changed.
  4. 4
    Reply with changes
    Point at what to change and where. Each reply edits the current version; earlier versions stay available.
  5. 5
    Export
    Export a full-resolution MP4 rendered from your original file.

A Prompt Formula That Works

Good editing prompts name four things: the material, the change, what must stay, and the result you want. You don’t need technical vocabulary; you need specifics.

Instead ofTry
“Make it better.”“Cut pauses longer than a second, remove the ums, and keep the pause before the punchline.”
“Fix the captions.”“The captions cover my face. Move them to the top and make them smaller.”
“Cut 0:12–0:47.”“Remove 0:12–0:47 of the current preview,” or “Keep only 0:12–0:47 of the original.”
“Add music.”“Put the track I uploaded under my voice, quieter while I talk, fading out at the end.”
  • Say keep or remove. “Cut 0:12–0:47” is ambiguous; “remove” or “keep only” is not.
  • Name the clock. Is 1:42 in the original recording or the current preview? After cuts they differ.
  • Batch related changes. “Cut the pauses, then caption it” lets the agent time captions to the final cut.
  • Protect what matters. Mention the caveat, the price or the name that must survive.

34 Example Prompts by Task

Each prompt below works as written. The tag after it is the editing tool the agent calls, by the name your own assistant sees over MCP. Music comes from a file you upload or a specific song you name; there is no stock music library or AI music generation, and sound effects are real recordings.

Cleanup

  • “Remove every pause longer than half a second.”cut_silences
  • “Take out the ums and uhs.”remove_filler_words
  • “I say the intro three times. Keep only the last take.”search_transcriptcut_range
  • “Cut everything before I say “let’s get started”.”search_transcriptcut_range

Captions and text

  • “Add captions with the active word highlighted in yellow.”add_captions
  • “Make the captions bigger and move them to the top.”set_caption_style
  • “It wrote “Valmira”. It should be “Valmera”.”set_caption_fixes
  • “Show “Chapter 2: Pricing” on a title card before the pricing section.”add_title_card

Framing and motion

  • “Make a 9:16 version and keep me in frame.”auto_reframe
  • “Zoom in slowly while I explain the chart.”add_zoom
  • “Speed up the setup section to 2x but keep my voice natural.”set_speed
  • “Go vertical just for the phone demo, then back to wide.”add_aspect_shift

Music and sound

  • “Put the track I uploaded under my voice, quieter while I talk.”add_music
  • “Use “[song name]” by [artist] as the intro music.”find_songfetch_urladd_music
  • “Fade the music out after the intro.”set_music_fit
  • “Cut to the beat of the song.”beat_align_cuts
  • “Add a whoosh on each scene change.”search_sfxadd_sfx

B-roll and media

  • “When I talk about the city, cut to b-roll of busy streets.”research_brolladd_stock_media
  • “Put my logo in the top-right corner for the whole video.”add_overlay
  • “Insert the clip I uploaded right after the introduction.”insert_media
  • “Use the video at this link as b-roll.”fetch_urladd_overlay

Look and finish

  • “Give it a warm, cinematic look.”apply_look
  • “Add a dip-to-black transition at each scene change.”set_transitions
  • “Fade in from black at the start and out at the end.”set_fades
  • “It looks a bit soft. Make it clearer.”enhance_video

Shorts and versions

  • “Find the three best complete stories and make each one a short.”make_shorts
  • “Cut this down to under 60 seconds, keeping the strongest part.”get_editorial_mapkeep_segments
  • “Now make a square version for the feed.”set_frame

Fixes and revisions

  • “The cut at 1:42 clips my last word. Give it back.”restore_range
  • “Remove 0:12 to 0:17 of the current preview.”cut_output_range
  • “Take the music out and keep everything else.”remove_music

Screen recordings and privacy

  • “Make the cursor bigger and steadier.”enhance_cursor
  • “Blur the email address on screen from 0:42 to 0:50.”blur_region
  • “Remove the burned-in subtitles from this clip.”erase_burned_text

Every tool is documented in the tool reference, and single jobs have their own guides, such as removing filler words, karaoke captions and beat-synced cuts.

How to Phrase Revisions

A good revision names the place, the problem and what to keep:

  • “At 0:18 in the preview, my last word is clipped. Restore it and leave a natural breath.”
  • “The music is too loud in the first ten seconds. Lower it there only.”
  • “Keep the captions and crop as they are, but swap the music for the other track I uploaded.”
  • “Remove the b-roll you added in the second half and keep everything else.”

Each change writes a new version and your original file is never modified, so nothing is lost. Studio’s timeline and editable transcript are there for the nudges that are faster by hand.

Describe Your First Edit

Upload a recording and type what you want. Creating an account and uploading are free; editing requires a subscription.

Create an account →
See pricing →

Chat vs Timeline: When Each Is Faster

TaskChatTimeline
Cleaning up a long recordingOne sentenceHundreds of manual cuts
CaptionsOne request, then style tweaksGenerate, then fix each line by hand
A frame-exact trimWorks with timestampsFaster with drag handles
Versions for several platformsAsk for each format and export itDuplicate the sequence and reframe each one
Nudging one cut by a few framesPossible, but wordyFaster by hand

Valmera gives you both: chat for the bulk of the edit, and a timeline for the last few frames. If you want the entire edit from a single message, see the autonomous video editor guide; for what makes the chat-driven approach work, see what an agentic video editor is.

Chat From Claude or ChatGPT Instead

If you already work in an AI assistant, connect it to Valmera’s MCP server and give it the same prompts. It calls the editing tools directly, renders previews and exports the MP4. Guides: edit video with Claude, edit video with ChatGPT, and using your own AI agent to edit video.

Other Ways to Edit Video With Natural Language

ProductHow written direction works
Descript UnderlordAn AI co-editor inside a transcript-based editor; prompts apply captions, clips, audio and visual edits
Premiere AI AssistantPrompts organize media and build rough assemblies on a Premiere timeline (public beta)
VEED OpenEditA coding agent that composes and renders edits locally on Apple Silicon Macs
CapCut Auto CutMusic- or speech-driven cuts, with text instructions in supported desktop and mobile versions

Checked against each vendor’s documentation on September 9, 2026.

Limits

  • Prompts edit footage you provide; they don’t generate a new video from scratch.
  • No true crossfades, motion-tracked stickers or custom font uploads.
  • Captions are burned in, with no SRT or VTT export.
  • English is the best-tested transcription language.

Frequently Asked Questions

Yes. In Valmera you upload footage and type what you want, such as “cut the ums and add captions.” An AI agent performs the edit on your real footage, renders a preview and reports back. You keep chatting to refine it, then export a full-resolution MP4.
No. Every edit can be requested in plain English. Studio also has a timeline and an editable transcript for precise nudges, so you can mix both, but you can finish a video without touching either.
A video editor you direct with written instructions instead of tools and menus. You describe the result, and an AI agent translates it into editing operations such as cuts, captions, reframes and music, performs them and shows you the result. It is also called natural-language or conversational video editing.
It follows supported requests and checks its own preview. When a request is ambiguous it asks; when something isn’t supported, such as a true crossfade, it says so and offers the closest option. If a result is off, describe what’s wrong and where, and it fixes that part.
Yes. Say “it heard Valmira, it should be Valmera” and the captions update everywhere the word appears. This changes the text on screen, not the recorded audio.
Yes. Connect Valmera’s MCP server (https://valmera.io/mcp/server) to Claude, ChatGPT or Cursor, and chat with your own assistant: it uploads, edits, renders previews and exports the MP4. MCP access is included with eligible subscriptions.
In Valmera, creating an account and uploading are free. Editing requires a subscription; see the plans page for current options.

Made with Valmera: real exports

Edit Your Video in Plain English

Upload footage, describe the edit, and review the preview. Creating an account and uploading are free; editing requires a subscription.

Create an account →
See pricing →

Related Articles

Valmera AI Video Editor
The product behind these prompts: capabilities, specs and limits.
Prompt Library
More editing prompts, organized by video type.
Autonomous Video Editor
Complete one-message briefs for a full rough cut.
How to Write AI Video Editing Prompts
The longer guide to briefs and revisions.
Timeline and Transcript
Versions, restores and hands-on adjustments.