A chat inside Premiere that understands your transcript and can use the whole plugin. You type a normal instruction and it works on your real timeline.
Sound familiar?You know what to cut, not where it is.
Scrubbing forty minutes of interview for the line where the guest talked about pricing is time nobody pays you for.
HyperChat works on the active sequence and loads its transcript on its own, Premiere’s or the plugin’s. You write a normal instruction and it carries it out, on your real timeline and with your clips.
Under the + button it also has five ready-made tools for what repeats most when you edit from a transcript.
One box for open requests, plus armed tools in the + menu for the edits you repeat.
Every button has its written command. HyperChat knows the actions of every tool and takes the values from what you type. It tells you what it will do before it does it, and why when something can’t be done.
One continuous cut of up to 60 seconds with the lines that tell the story. Along the way it removes silences, filler words and repeated words, keeps the last clean take of each idea and only the interviewee’s voice. With or without a hook.
For interviews, podcasts and long videos with several topics. You get a table with a suggested title, duration, ranges and the transcript corrections it found. Approve, and every cut lands in Premiere with its title as a marker. Each clip brings its hook, ready to insert at the playhead. Download the script as .xlsx or .txt, and import the .xlsx later to rebuild the videos without spending AI again. It uses the reels-cutter skill that comes with the plugin.
Name the speakers, say which camera shows whom and choose a pace. HyperChat cuts the camera tracks so you always see whoever talks. Audio untouched, or one mic per speaker cut with the camera.
A quick list of where a shot or a graphic would help, and what kind. It recommends; it does not download footage.
The transcript loads itself. Attach is for a .json transcript you exported or an .xlsx script you made earlier.
Claude through your Claude Code subscription, GPT through Codex, or any of them with your API key. You set it up once in Settings.
Three real jobs, with the buttons named as in the panel. Each animation shows what the timeline does.
A half-hour talk, and the client wants one minute of the best.
Press + › Summarize. The box fills with its mini prompt and leaves “with hook” or “without hook” selected for you to type over.
With hook, the strongest line opens the cut and the rest follows in order. Without hook, it keeps the chronological order. Add a tweak if you want, for example “up to 90 s”, and press Enter.
Check the cut in the chat: silences, filler words and repeated takes are already out. Build it in the sequence.
“The first time I got paid for a video, I charged almost nothing. I learned that the result sets the price, not the hours. Today I send the proposal before I open Premiere. And the client decides with the value in front of them.”
Illustrative animation. Pause it with the button at the bottom right.
A one-hour podcast has ten reels inside. Finding them, titling them and cutting them by hand takes a whole afternoon.
Press + › Script and choose how many clips: All, 3, 5 or 10.
For example “5 clips up to 45 s about pricing”, then press Enter. Sort the table by virality or by position and check the ones you want.
Approve all and build them: each cut lands in Premiere with its title as a marker. It also works on a multicam if cameras and audio are synced.
Download it as .xlsx or .txt. If you import the .xlsx later with Attach, HyperPowers rebuilds the videos without spending AI again.
Illustrative animation. Pause it with the button at the bottom right.
Three cameras, two guests, an hour of conversation. Switching angles by hand on every turn is the most boring part.
Sync the cameras and transcribe the sequence in Premiere, with speakers. The multicam relies on that transcript.
Press + › Podcast · Multicam. Name the speakers, choose the track of each camera and who appears on it, and choose the audio: one audio for everything, or one mic per speaker with its track.
Choose the pace and press Analyze, then Cut. Undo cut brings everything back, audio included.
Press + › B-roll & graphics for a list of the shots and graphics that would help, by moment. Add “graphics only” or “five at most” if you want.
When I started, I edited at night in the kitchen at home.
Today I follow three steps before opening Premiere.
And the most important thing is talking to the client before quoting.
Illustrative animation. Pause it with the button at the bottom right.
HyperChat finds the moments. The other tools finish them.
Cut the reels with Script, then caption each one with Smart groups that never split a phrase.
Cut by text, then clean the voice and tighten the remaining pauses on the waveform.
Cut the multicam podcast, then give each speaker a box for the vertical version.
Same panel, same license. Every tool works on its own and hands off to the next.
Captions you don’t have to redo, and text you can animate by hand.
03Animated captions inside Premiere, line by line and editable.
04Clean the voice, split the music, cut silences on the waveform.
05Make text follow whatever moves, or stabilize a shot.
06Split screens from your clips, without doing the math.
07The one-click buttons Premiere is missing.
08Your own sound library, placed right on the hit.
09Cinematic color, guided, loaded into Lumetri.
Nine tools.
One license.