← Home
LANDING

Published · Updated

Edit Video by Chatting With AI

No timeline. No tracks, no keyframes, no dragging clip edges. You upload footage, type what you want changed, and an AI agent performs the edit and shows you a preview — then you react to it the same way you would react to an editor who just sent you a cut.

Below is a representative exchange, corrections and all, ending in a downloadable file.

Have This Conversation About Your Own Footage

50 free credits on signup, no card required.

Start free →
See pricing →

What the Conversation Actually Looks Like

YOU

Cut all the dead air and the ums out of this.

VALMERA

Removed 47 silences and 31 filler words, all snapped to word boundaries. Your 12:04 recording is now 9:18. Preview below — the pacing through the middle section is much tighter, though the pause before your closing line got cut and it might have been deliberate.

YOU

Yeah, put that pause back. And add captions.

VALMERA

Restored the 1.4s pause before "and that's really the whole point". Added word-accurate captions in the podcast preset — white text, bottom third, four words a line.

YOU

Captions are too small and I want the key words highlighted.

VALMERA

Bumped the captions to large and turned on per-word emphasis with a yellow highlight. Restyled in place, so the timing is untouched.

YOU

Put some chill music under it, quiet, and make it feel a bit more energetic overall.

VALMERA

Added a chill track at bed level with ducking, so it drops under your voice and comes back in the gaps. For the energy: punch-ins on your eight most emphasized words, and the cuts now use a whip transition at the three real scene changes — not at every jump cut, since those are inside one continuous take and are meant to be invisible.

YOU

Make it vertical for Reels and export it.

VALMERA

Reframed to 9:16 — sampled frames across the edit and aimed the crop at where you actually are rather than the centre. Captions rescaled to the new frame. Exporting from your original file at source quality now; I will hand you the download link when it lands.

Illustrative of the interaction and the kind of detail the agent reports back; your footage and its numbers will differ.

Why the Corrections Are the Point

Notice what happens in the second turn above: a pause gets restored without disturbing the other forty-six cuts. In a timeline, changing your mind about one decision made three steps ago is expensive — you scrub, you find it, you hope nothing downstream has shifted. Here it costs one sentence, because the agent edits a decision list rather than pixels and every operation is a reversible version.

That is what makes conversational editing viable rather than merely novel. The interaction is only pleasant if being wrong is cheap.

How to Edit a Video by Chatting

  1. 1
    Upload and let it read the footage
    Up to 14GB or 3 hours. A one-time analysis produces a word-level transcript, the silences, the shot boundaries and descriptions of what is on screen, with visible progress. Every later edit reuses it.
  2. 2
    Say what you want, in your own words
    "Cut the boring parts and add captions" is enough. You can stack several dependent requests into one sentence — the agent sequences them itself.
  3. 3
    React to the preview
    Watch what comes back and correct it conversationally: "put that pause back", "bigger captions", "quieter music". Then ask it to export, and download a source-quality MP4.

A long upload takes a while to analyze and shows progress throughout; after that first pass, edits are conversational.

Or Chat in an Assistant You Already Use

The same engine is published over the Model Context Protocol, so you can have this conversation inside Claude instead — the model in your session picks the tools, Valmera performs the edit, and you get the same downloadable file. Setup is one paste and one sign-in.

Frequently Asked Questions

Yes. You upload real footage, it is analyzed once into a word-level transcript, detected silences, shot boundaries and visual descriptions, and then you type what you want changed. An AI agent performs the edit, renders a preview, and you react in plain English — "looser cuts", "bigger captions", "different track". You never open a timeline unless you want to.
No. "Cut the boring bits", "make it punchier", "the music is too loud" all work. The agent maps everyday language onto real operations. Knowing the vocabulary makes you faster — asking for "a J-cut" or "−14 LUFS" gets you exactly that — but it is never required.
There is one, and you can ignore it. Valmera has a visual timeline, a preview player, aspect-ratio controls and an editable transcript for when you want to nudge something by hand. The difference from a conventional editor is that none of it is required — a whole video can be finished without opening any of it.
You say so, and it changes. Every operation is a version, so corrections are cheap and non-destructive — restoring a cut it should not have made does not disturb the other forty. This is the real advantage of the conversational form: in a timeline, a change of mind three steps back is expensive, and here it is one sentence.
The test is whether one sentence containing several dependent steps gets carried out. "Cut the dead air, caption it, put music under my voice, and reframe for Reels" involves cuts that change the timeline the captions must be timed against, and music that has to fit a program length nothing knows until the cuts are done. An agentic editor sequences that itself. A chat panel bolted onto a manual editor typically triggers one feature at a time and leaves the sequencing to you.
No — every reply is verified server-side against the edit decisions actually recorded. The agent cannot claim a change it did not make, and when nothing changed it says so. That matters more in a chat interface than anywhere else, because you are trusting a written report rather than watching your own hands.
Yes. Valmera publishes its complete editing toolset over the Model Context Protocol, so Claude can drive the same engine from your own conversation — upload, cut, caption, mix, render, export and download. Setup is one paste and one sign-in.
A downloadable H.264 MP4, rendered from your ORIGINAL upload at source quality — previews use a faster proxy, but the export never does. Free-plan exports carry a small Valmera mark in the corner; every paid plan exports with no watermark. Every export closes with a brief (~2.5-second) end card after your content.

Type the Edit Instead of Making It

50 free credits, no card required. Upload something and tell it what is wrong with it.

Start free →
See pricing →

Related Articles

Agentic Video Editor
The category: what it means for the agent to be the editor.
Claude Video Editor
Drive the same engine from your own Claude conversation over MCP.
50 Video Editing Prompts
Real requests that work, copy-pasteable.
Text-Based Video Editing
Editing by editing the transcript, for when you want hands-on control.