PowerCut User Guide

Getting started

Requires macOS 26 (Tahoe) for transcription. On-device transcription uses Apple's speech engine, which exists only on macOS 26 and newer. On macOS 14 and 15 PowerCut opens and edits existing projects (the verbatim engine still runs) but cannot transcribe new footage. An Apple Silicon Mac is required.

Languages. PowerCut speaks twenty languages, chosen per project or per clip: see Languages.

The idea. PowerCut is the assistant editor before your real editor. It takes raw talking-head footage, transcribes it on your Mac, produces a clean first cut automatically, lets you polish that cut by editing text, and exports a frame-accurate timeline to Final Cut Pro, DaVinci Resolve or Premiere Pro. It is not an NLE and does not want to be one: no effects, no color, no titles. It does one job, fast.

Take the tour. On first launch PowerCut offers a guided quick start: fourteen short steps that ride your own first project, each pointing at the real control, from dropping in footage, adding more later and building a multicam clip, to export. Skip it at any moment, go back a step any time, and replay it whenever you like from Help → Take the Tour. The rest of this manual is the long version.

The layout. The toolbar carries the project: Home and the Footage toggle at the left, then the project's name in an outlined field (click it to rename) with the project menu right beside it (new, open and save projects, and your recent ones). At the right: History, Review with the number of close calls, Export, the toggle for the right column, and the gear for Settings. Below the toolbar, Footage sits on the left, the transcript (the paper edit) in the middle with the timeline under it, and the video preview on the right with three tabs under it: Edit (the rules the Auto Edit Pass follows), Review and Assistant. The divider between transcript and preview is draggable, and a thin status bar at the bottom shows the format of the clip under the playhead. Longer explanations sit behind the small (i) buttons: click one for the details.

Home. The house in the toolbar (⌘⇧H, or View → Home) shows every project you have made, as a list or a grid, each with a frame of its footage so you recognise it at a glance. Click one to open it (the current project autosaves first). The first tile, New Project, starts a new one (so does New Project in the toolbar), and Open… opens a project file from anywhere. The column at the right has the tour, this guide and the keyboard shortcuts one click away, and your trial or license status when it matters. Back, or Esc, returns you to where you were. The New Project page lists your previous projects too, and has its own Back link. The frame is taken from the first clip and saved beside the project file, so the list is instant and works with the drive unplugged; audio-only projects show a waveform instead.

Projects save themselves. Moments after your first import, a project file exists in ~/Movies/PowerCut (changeable in settings), and every edit autosaves a couple of seconds after you make it. Name the project in the name field at the top left of the toolbar; the file renames itself to match. Use File → Save Project As to keep a project somewhere specific; it autosaves there from then on.

Languages

PowerCut transcribes and cuts in English, Spanish, French, German, Italian and Portuguese, plus Japanese, Korean, Chinese (Simplified and Traditional), Cantonese, Hindi, Hinglish, Vietnamese, Russian, Ukrainian, Polish, Dutch, Swedish, Lithuanian and Greek. Everything non-English is labeled beta: transcription and filler detection keep improving there, and if something reads wrong in your language, tell me, real clips are what tune them. The language drives everything downstream: the transcription engine, the verbatim pass that hears filler words, and the filler vocabulary itself (um and uh in English, euh in French, äh and ähm in German, えーと in Japanese, and so on).

What beta means per language. Spanish, French, German, Italian, Portuguese and Japanese run the full pipeline, filler and stutter detection included, on Apple's transcriber. Russian, Ukrainian, Polish, Dutch, Swedish, Lithuanian and Greek also run the full pipeline, filler words included, with the clean transcript carried by the on-device Whisper engine instead (Apple's transcriber does not offer them yet). Korean, Chinese, Cantonese, Hindi, Hinglish and Vietnamese run on a clean transcription engine alone for now: transcription, pause cuts, takes and the whole paper edit work, but there is no filler-word pass yet. The English-specific cleanup rules (production chatter, spoken directives) apply only to English.

Whisper-carried languages (Russian, Ukrainian, Polish, Dutch, Swedish, Lithuanian, Greek, Hindi, Hinglish, Vietnamese) download the Whisper model on first use, about 1.5 GB, which can take several minutes; after that it is fast and entirely local. Accuracy varies by language, strongest for Russian, Polish and Dutch, and these are the newest betas, so real clips are exactly what tunes them: if something reads wrong, tell me. Hinglish uses Whisper's Hindi mode, built for that Hindi-English mix; expect output to lean toward Devanagari script. Greek reads its question mark, the semicolon, as a sentence end when a Greek word comes before it (a question ending in a number or a Latin name is not caught); write a Greek script in normal lowercase Greek, since an all-caps script drops the accents and the final sigma and can fail to match the transcript.

Translate to English (beta). Any language can be transcribed and translated to English on device: Whisper listens to the spoken language and writes the transcript in English, with every word timed to the original footage, so cuts land exactly where things were said. Choose Translate to English… at the bottom of any clip's language menu, or the Translate… button under Run Auto Edit Pass for the whole project. In the dialog, name the spoken language (a touch more accurate) or leave Auto Detect for mixed or unknown footage; translation always writes English, that is the one language Whisper translates into. If the clip is already edited, Keep my current edit carries your strikes and cuts onto the English words by time, so the cut does not move; or run the Auto Edit Pass fresh on the translated transcript. Wording corrections you made to the old words do not carry over, they described text that no longer exists. Nothing is touched until the translation actually succeeds: a failed download or a quit mid-translate leaves your project exactly as it was. Translation uses a translation-capable Whisper model (about 1 GB, downloads once), and an English script will guide take picking on translated footage.

Transcription accuracy. In PowerCut's Settings (and the Edit tab under the preview, and the Translate dialog, they are the same switch) you can choose Standard, the model recommended for your Mac, Maximum, which runs OpenAI's full Large-v3 model (noticeably more accurate on Hindi and on translation), or Large v2, the older full model, which is steadier than v3 on Cantonese and Chinese. Maximum and Large v2 are each about a 3 GB one-time download and slower to run. The choice applies to the Whisper-carried options (Hindi, Hinglish and translation) from the next transcription on.

Cantonese and Chinese: two engines to choose from. At Standard, Cantonese, Simplified and Traditional Chinese transcribe on Apple's engine. Set accuracy to Maximum or Large v2 and they move to Whisper instead, on that model. If a clip comes back empty or as one sentence repeated over and over (Cantonese does this on some clips), PowerCut retries it once on another large Whisper model that is already on your Mac, never a download, and keeps whichever read is usable. Cantonese also has a script choice in the Edit tab, shown when a Cantonese clip is in the project: Traditional (Hong Kong's, the default) or Simplified. Changing it transcribes the Cantonese clips again in the new script.

Auto Detect for mixed-language clips (beta). At the bottom of every language menu. Whisper re-detects the spoken language about every 30 seconds and writes each part in its own script: one clip, several languages, one transcript. Best for footage that genuinely switches language mid-recording; brief interjections inside a stretch of another language lean toward the surrounding language. Transcription and pause cuts work; the filler-word pass isn't available in this mode.

The project's main language. Pick it in step 1 of the New Project page, before you import. Every clip transcribes in this language unless you override it. Changing the main language later only affects clips you add afterwards; clips that are already transcribed keep their words and their language.

It remembers, and it listens. A new project starts in the language you picked last, so a channel that always shoots in one language never touches the picker again. And unless you picked a language yourself, PowerCut listens to the first thirty seconds of the first clip and sets the project language from what it hears, before transcription starts; the Footage panel shows "Language detected". Detection uses the Whisper base model: if it is not on your Mac yet, the Footage panel offers a one-time download instead of fetching it silently. Turn automatic detection off under Settings, Import.

Per-clip language. A bilingual shoot is just two clips with two languages: every clip card in the sidebar shows a globe dropdown with the clip's language (right-clicking the card offers the same menu under Clip Language). Changing a clip's language re-runs its automatic pass from scratch, so text corrections and strikes on that clip are redone fresh; PowerCut asks before it does this.

Voice Isolation, per clip. Footage from outdoors, in wind or in a crowd transcribes worse and its pauses are harder to find. Right-click the clip and choose Voice Isolation Low, Medium or High: PowerCut runs Apple's Voice Isolation on the audio it listens to before it analyzes and transcribes the clip, and the card shows the level. Playback and the export keep the original audio untouched; this only changes what PowerCut hears. The clip re-runs, and the transcript and edits for each level are remembered, so switching back is instant. Save Project with Audio keeps the clip's own recorded audio: isolation is applied again when the package is opened.

First use of a language downloads its models. The system speech model and the verbatim engine's multilingual model fetch once, then live on your Mac; expect the first non-English clip to sit in "transcribing" for a while before words appear. After that, it is as fast as English, and nothing is ever uploaded.

Need a language that is not here? Email apps@rafaelludwig.com; requests decide what ships next.

Import & transcription

Drag video files into the window, or click Add Footage. Import as many clips as you like; drag the clip cards to reorder them, and the story follows that order. Each card shows a frame of the clip, its codec, frame rate, resolution, audio layout and any warnings (variable frame rate, missing audio, and so on). To fit more clips, hide the frames with the photo button in the Footage header, or View → Hide Clip Previews; each card then shows its length beside its name.

Transcription starts automatically and streams in: you can read, play and edit while long files are still processing. Expect roughly 20 seconds for 20 minutes of footage on Apple Silicon.

When footage goes missing

Unplugged the drive, or moved a file? The clip stays in your project, marked red with MISSING FILE, and every edit and transcript keeps working from saved data. Playback holds the clip's place as silence, and exports still reference the original file so your NLE can relink. Plug the drive back in and everything heals automatically; for moved files, right-click the clip and choose Locate Missing File.

Importing a multicam or compound clip from Final Cut Pro or DaVinci Resolve

Already built a multicam or a compound clip in your NLE? Export it as FCPXML (File, Export XML in Final Cut Pro; a .fcpxml file or a .fcpxmld bundle both work) and bring it in with File, Import Final Cut Pro / Resolve XML, or drop the file anywhere in the window. PowerCut reads the clip's structure, not the edit: the primary angle's footage becomes the clip, every other angle becomes a linked angle at its exact offset (open Angles to see them; blank angles keep their slot), and the dialogue roles in the XML decide the audio: a dialogue channel on the primary selects its track, and every other angle tagged as dialogue is mixed into the same transcript at its offset. A two-speaker interview recorded as one file per person therefore imports as one clip whose transcript has both voices, with nothing to click. A summary lists what arrived, and any footage that isn't reachable comes in offline with its duration and format from the XML, ready for Relink.

The angles leave with the cut. Importing a multicam sets that project's export to Multicam For Each Clip (the Export sheet says so, and remembers it per project; Use My Default returns to your app-wide choice). The export rebuilds one multicam per source under the names it came in with, the container's name and the angle names, with your cuts applied as multicam clips and every angle whose audio was on in the imported timeline, or that you linked as dialogue, switched on again. It is a rebuilt multicam, not the original: the container, the angle order and the internal ids are new, and Multicam All Clips exports carry no angles. Premiere receives the angles on stacked tracks (see Exporting).

Choosing what to transcribe. The import summary shows every audio source the XML brought in, one row each: the main clip's tracks and every angle that has audio. Tick the rows the transcript should hear; several rows are mixed and read as one signal. The boxes start the way the XML describes the dialogue, and Reset to XML puts them back. A second column, Hear, keeps a row audible in the player without transcribing it, for a computer-audio or music track. A compound audio angle appears as a preset that ticks its members. After the import the same boxes live in the clip's audio menu under Transcribe from and Hear in playback only, so what you chose is what you see, and you can change it any time: PowerCut re-runs the pass for a new combination and remembers the transcript and edits of each one.

One file per angle lands as exactly one clip. A container whose angle holds several files back to back (a whole shoot in one multicam, or PowerCut's own All Clips exports) imports each file as its own clip, in order, with the angles linked per clip. The cuts in the exported project are not imported; that is what PowerCut is for.

Supported formats

RED RAW (R3D) BETA

PowerCut cuts by what was said, so it does not need RED's picture to do its work. On import it reads the sound the camera recorded into the R3D (a one-time step of a second or two, cached afterwards) and takes the picture's size, frame rate, length and start timecode from the file's header. From there everything works as with any clip: transcript, Auto Edit Pass, takes, the timeline. The player stays dark for these clips and says so; the picture appears when the cut opens in your NLE.

• Exports reference the original R3D, which Final Cut Pro, DaVinci Resolve and Premiere Pro read natively. If the clips come in offline, select them and relink to the clip's .RDC folder (Resolve: right-click, Relink Selected Clips). The App Store version of Resolve can only open media you have pointed it at, so it needs this step.
• The clip needs camera sound. A RED clip recorded without audio has nothing to transcribe; sync an external recorder in your NLE and cut from a proxy instead.
• Not yet: an R3D as a secondary multicam angle, and the older RED ONE era files. If PowerCut says it could not read a clip's sound, make a proxy (ProRes or H.264) in your NLE or REDCINE-X and import that; the cut relinks to the R3D by name and timecode.

Two engines, one transcript

PowerCut listens twice. Apple's engine produces the clean transcript you read. A second verbatim engine (a one-time ~500 MB download) hears every um, uh and stutter that the clean engine smooths over. The two are cross-checked: filler words are cut only where both engines agree, invisible doubled words ("this… this") are caught even when the clean transcript shows one word, and words the engines disagree about get a dotted amber underline so you proofread five words instead of five thousand. When an amber word is right as it reads, press Y (or right-click it → Looks Right, or use the Looks Right button in the correction popover): the underline goes, the word stays settled, and the selection jumps to the next amber word, so a proofreading pass is Y, Y, Y. Edit → Next Uncertain Word (⌥⌘→) steps without confirming. Correcting a word settles it the same way.

Multi-track audio

Screen recorders and field recorders often carry several audio tracks. If a clip has more than one, its card shows "track 1/3"; right-click the card → Audio Track for Transcript to switch. PowerCut re-analyzes from exactly that track, and playback uses the same track, so what you hear always matches what was transcribed.

Several tracks as one voice. A multicam export or a recorder with one mic per track puts each speaker on a different track. In the same menu, the section Read together as one signal ticks any number of tracks into one combined signal: transcription, the pause analysis and playback all hear everyone, mixed at equal level. The label reads "Audio: Tracks 1 + 2". Changing the set re-runs the clip, or restores the transcript it already has for that combination.

The paper edit

The transcript is the edit. What is struck through is out of the cut; everything else is in. The timeline, the preview and the export all follow the text, instantly.

• Click a word to place the caret and move the playhead there. Double-click to correct a word's text (audio and timing never change). The correction popup can also replace every matching phrase in the project in one step, and remember the correction so every future transcript applies it automatically. Cancel or Escape closes without changing anything.
• Find anything with ⌘F: words or phrases across every clip, close matches included, so a mis-transcribed name still turns up. Press Find for a results list with timecode, clip and match percentage; every match glows amber in the transcript, and clicking a result jumps there.
• Drag across words to select. A floating bar appears: strike, edit, copy or cut. Select several words and edit them as one phrase.
• ⌘F finds words or phrases anywhere in the transcript, across every clip. Close matches count, so a mis-transcribed name still turns up. Return jumps to the next hit.
• S strikes or restores the selection (or the word at the caret) in one key; U always restores. ⌘⌫ and ⇧⌘⌫ do the same from the menu. While you are correcting text, shortcuts stay out of the way.
• Return splits the paragraph (and the clip) at the caret. Backspace at a paragraph start merges it with the previous one.
• Drag a selection to move it: words, sentences, whole beats reorder, and the video reorders with them. ⌘X/⌘C/⌘V work the same way, in the transcript and the Document view alike: a blinking caret shows exactly where pasted words will land.
• Edit and Document at the top of the transcript switch between this editing surface and a clean reading page (next section); the (i) beside them lists the mouse and key moves, and the A buttons set the text size.
• The Hide Cuts toggle in the transcript header gives you a clean read: struck text collapses behind faint ✂ marks, fully struck passages fold into a single hairline (right-click either to restore), and the timeline collapses the same strikes. The timeline's own Hide Strikes toggle can override the timeline side independently. Either way, struck words are skipped in the cut.
• The Speakers switch in the transcript header shows or hides the speaker names on a project with labeled speakers (see Multicam angles); the names never affect the cut.
• Clip boundaries show as a full-width rule with the clip's name. During playback the current word glides under a teal highlight, and the transcript follows along: when the spoken word reaches the bottom of the panel, the page turns so it stays in view, the way the timeline follows the playhead. Scroll or scrub yourself and it waits until you stop. The follow toggle sits in the transcript header next to Hide Cuts.

The Document view

Document, beside Edit at the top of the transcript, turns the transcript into a clean paper page: serif type on white, timestamps in the margin, a heading whenever the clip changes, and a zoom slider (or ⌘−/⌘=/⌘0) that scales the whole document. It is built for producers and paper edits: reading, searching and quoting the footage like a script.

• Read mode is the default: the fastest way to scroll and search a long transcript. Click any word and the video plays from that moment; drag to select, then play or copy the selection from the floating bar. During playback a highlight reads along word by word and the page follows.
• Edit Document (the button above the zoom slider) mounts the full editing surface on the page: strike, restore, correct, and drag material to a new spot, exactly like the chip view. Clicking a clip in the footage column jumps the document to that clip's section.
• Skip Cuts (the button above the zoom slider, on by default): playback in the Document view runs the edit, so only unstruck text plays and the read-along jumps over cut material. Turn it off to play the full footage, cuts included, for instance to audition a struck take by clicking into it. It is the same switch as Skip Strikes on the timeline.
• Three ways to see the cut: Strike Thru Cuts shows the full transcript with cut material struck through; turn it off for clean plain text; Hide Cuts collapses the cuts for a clean read. With the full transcript showing, playback switches to the full footage automatically, so every line plays when clicked, cut or kept. Material you moved shows a soft tint until you clear it.
• Word export, three flavors: Transcript… in the Export sheet (Export in the toolbar), the File menu, and the document's own Export menu writes a .docx: the full uncut record, the edited transcript with cuts as real strikethrough, or only what made the video edit. Reading- sized blocks with timestamps and clip headings, readable by Word, Pages and Google Docs.

The Auto Edit Pass

The New Project page, in three steps. Step 1 is about the shoot: name the project if you like, add the script if you have one (its words help recognition, and take selection follows it), tick Multiple speakers if more than one voice talks, and pick the language. Step 2 is the video style with the first-pass switches beside it; the style sets their defaults and every one stays adjustable. Step 3 is the footage: Choose Files, Create Multicam, or drop files anywhere on the page. Clips land in a list you can drag into order or sort, each with a Noisy checkbox (Voice Isolation at Medium on the audio PowerCut listens to; the header box ticks them all). Clips are analyzed and transcribed the moment they arrive, in the background, so by the time you have chosen a style the first ones are done; the row shows probing, analyzing, transcribing and transcribed. Changing the language, the microphone or Noisy transcribes that clip again. Then one button at the bottom, Analyze and transcribe (it reads Open the edit once every clip is transcribed): it applies the style, the checkboxes and the script to every clip at once and opens the editor; a clip still transcribing finishes there and takes its first pass when its words land. A project quit mid-setup reopens on the page with everything it had; a project that already ran opens in the editor.

Video styles. The style, chosen in step 2 of the New Project page or in the Edit tab under the preview, decides how the pass reads the footage. YouTube · Tech is the tight baseline: pauses cut hard, a tiny beat at sentence ends, every cleanup rule on. YouTube · Tutorial keeps generous pauses, keeps the pause after "like this" or "click here" for the demonstration, and never strikes an instruction that trails into one. Music Tutorial is for voiceover over music and playing: pauses filled by music on another channel or track stay, a phrase played into the voice mic stays, demonstrations stay ("listen to this", "sounds like this"), an instruction that trails into playing is never struck, pauses under 1 s stay at Natural, and the preview plays the whole track. YouTube · Storytelling keeps every sentence-end pause under 0.8 s for the beat and treats a line repeated for effect as intended, not as a re-read. Documentary keeps pauses under 1.5 s at sentence ends and 2 s mid-sentence, keeps a laugh or a reaction that has no words, keeps breaths, and does not strike thinking aloud. Vlog is brisk but treats "Nice!" and "Oh my god" as the content, never as chatter. Podcast / Interview keeps the conversation: only silences over 5 s go (6 s mid-sentence), laughter and reactions stay, breaths stay, and small talk, trail-offs and re-reads are not struck. In every style the filler-word and repeated-take checkboxes mean what they say. Picking a style on the New Project page also sets four cleanup knobs to that style's defaults: Podcast / Interview turns repeated takes, production asides and context judgment off at Light, Documentary keeps asides on with takes and context judgment off at Light, and every other style starts with all three on at Natural. Silences and filler words stay your own choice, the checkboxes remain adjustable right there, and in an open project a style switch leaves them alone; the Cleanup card's Style defaults button applies them on request, one undo step away. The cleanup level scales the pause floors (Light raises them, Balanced halves them, Aggressive and Jump Cut remove them), and the Pauses card shows the exact bars in use. Switching styles later re-tunes timing at once; the rule differences apply when you choose Also Run Auto Edit Pass in the switch dialog.

The pass runs automatically when a clip finishes transcribing, and on demand from the big Run Auto Edit Pass button in the sidebar (a version of the current edit is saved first, see Versions). It strikes what an experienced assistant editor would strike:

• Silences, tuned per video style: a Tutorial keeps natural demonstration pauses, a Podcast keeps conversational ones. The five levels under the cleanup switches set how eager the pause cuts are: Light takes only clear dead air and leaves every breath, Natural keeps breathing room and intentional pauses (the recommended level), Balanced tightens while still protecting dramatic pauses, Aggressive cuts everything cuttable, and Jump Cut, for talking heads cut tight, cuts every pause past the slider, mid-sentence too, down to 0.1 s, while keeping short words like "a" or "I" that sit between two cuts. Padding sliders control the air around every cut, including take and filler cuts. Changes apply when you release the knob (with a confirmation you can turn off), and each slider panel has a Reset back to the stock values.
• Music and other audio: PowerCut listens to the channels and tracks the transcript does not use (the other channel of the voice track, a music or system track). A pause in the voice that music or another voice fills on them stays in the cut whole and lists in the review queue as "Music or other audio kept" with a one-click Cut. Remove detected music in the Cleanup card, or on the New Project page, cuts those pauses like any other. The preview plays those channels at the level of the Other channels in preview slider in the Pauses card (Off at the left); the export always carries the whole file. The trims at a clip's very start and end are not protected. For this to work the footage row's audio choice has to isolate the voice: a track that holds voice on one channel and music on the other needs Left only or Right only, not Both.
• Demonstrations: when the words before a pause announce that something is about to happen on screen ("like this", "like so", "let me show you", "watch this", "click here", "drag it over here", "take a look"), the pause that follows is the video, not dead air, and it stays in the cut whole. On by default for the YouTube Tutorial and Tech styles; the Keep demonstrations checkbox in the Cleanup card turns it on for any style or off. Every kept demonstration lists in the review queue with its cue and a Cut button, and each one you cut teaches the rule to hold back on your footage.
• Background noise tolerance, the second slider in the Pauses card. PowerCut measures each clip's own room tone and sets the silence threshold above it automatically. If a fan, a hum, or street noise keeps pauses from being cut, move the slider right: faint noise then still counts as a pause. Move it left when quiet speech is being mistaken for silence. Zero is the automatic threshold; the plan updates when you let go, without re-analyzing the audio.
• Filler words (um, uh, and their equivalents such as euh and hum in French) and invisible doubled words, verified by both engines. An "uh" the transcript spells "ah" is cut when PowerCut also heard a hesitation there, while a real "Ah, I see" stays. For a project transcribed before 1.0.25, click Listen Again for Fillers in the Cleanup section to find them.
• False starts, restarts and re-reads: "Now, that's, now that's clean" keeps only the clean read. Parallel clauses that continue a thought ("I know you heard it often, and I know you didn't believe it") are read as one sentence, not as a repeat.
• Trail-offs (sentences that stop mid-thought) and production chatter ("Okay." "Sorry, again.").
• Production asides ("Why is my face red?"), caught even without a script by their isolated vocabulary.
• Repeated takes, grouped with the best read kept (next section).
• Optionally, an on-device context pass (Apple Intelligence) reads the cut like an editor and flags lines that are dead in context. Conservative by design; every flag shows a sparkle and its reason.

Repeated takes

When you read a line several times, PowerCut clusters the attempts into a take group. The TAKE 2/3 tags above paragraphs show each read's place in its group; the read in the cut shows a teal star and Take 2 of 3 instead. The pick prefers the final complete read, uses your script's coverage when one is loaded, breaks ties by delivery cleanliness, and can stitch the best start of one read to the ending of a pickup (badged START and ENDING).

• Hover a badge to halo every read in the group, so you can see them scattered through the transcript.
• Click a badge to use that read instead. Right-click for skip and re-apply options.
• Saying "use the first take" or "use the second take" on camera is honored as a spoken edit direction.
• A redone run of sentences inside one paragraph is found too: the retake becomes a take of its own, and the last full attempt is the one recommended.
• A short line said again word for word ("I'm sorry, but it does.") is a take too, however few words it has, as long as it is at least four words and the next copy starts within 30 seconds of the last. A leading "And", "But" or "So" does not make it a different line. For a project edited before 1.0.25, click Run Auto Edit Pass once to apply it.

The review queue

The Review button in the toolbar counts the decisions the automation was least sure about: close-call take picks, context strikes and possible mishears. Click it, or the Review tab under the preview, to open the list. That honesty is the point: check nine flagged calls instead of proofreading everything.

• Click a row (anywhere) to jump the transcript and playhead there; the paragraph halos so you can see exactly what is being judged.
• Play auditions the moment in place and stops at its end. For take picks, numbered buttons play each read, including the skipped ones.
• Resolve with one click: keep the pick, restore the line, or confirm the text. Struck material auditions through Full Footage automatically and your preview mode comes back when you leave the Review tab.

Timeline & trimming

Three controls shape the timeline. Rough Cut shows the edit: struck material greyed out with a reason stripe (grey pause, indigo strike, amber filler, red manual), or collapsed away entirely with Hide Strikes so only what ships is on the strip. Full Footage shows each file in the Footage panel as one solid clip, with no edits drawn and no edit actions. Skip Strikes (on by default, beside Hide Strikes) is the playback switch in every layout: on, playback jumps over struck clips; off, it plays everything, and in Full Footage that means the whole files. Follow keeps the word that is playing, or the word you click in the transcript, in the middle of the timeline at any zoom; switch it off and the timeline stays where you put it. Follow only moves the view while something is playing or when you jump somewhere new: after a trim the timeline stays exactly where it is, so you can see the edit you just made.

The clock beside the transport reads timecode at the project's frame rate (drop-frame for 29.97 and 59.94), the ruler draws single frames once you zoom in far enough, and every trim lands on a whole frame, with a "+N fr" readout while you drag.

More room. The Footage panel on the left and the preview column on the right can each be put away: View → Hide Footage Panel (⌘1), View → Hide Player and Settings (⌘2), or the two panel buttons in the toolbar. The transcript and timeline take the space, playback keeps running, and PowerCut remembers the choice. With the Footage panel hidden, analysis and Auto Edit progress show above the transcript.

• The ruler scrubs; the clips select. Click a clip to select it (the transcript jumps along), drag across clips or shift-click a later clip to select the whole run between, then Strike, Restore, Remove or Delete.
• Skim plays audio under the pointer while paused, so you can find a word by ear on the waveform.
• Trimming: every clip has grip bars at its edges; the yellow bracket appears as you hover, FCP-style. Drag inward to tighten, outward to reveal, with the actual incoming waveform drawn live and downstream clips rippling as you drag. Trims override strikes at frame precision; expose most of a word and its text un-strikes to match.
• Range tool (R): drag a span on the waveform, press Delete, and it's gone. The surgical fix for a stray noise or syllable.
• Pulling a clip past its neighbour: when the next clip continues the same footage, a plain outward drag rolls the edit: this clip grows, the neighbour gives up the same frames, and nothing plays twice. Hold ⌥ while dragging to overlap instead: the clip grows as far as its file allows and the shared part plays in both clips. Shared material is drawn with a faint diagonal hatch in both.
• A repeated section is edited on its own. After an overlap, the pulled part starts as a true copy; from then on deleting, trimming or striking it in one clip leaves the other alone, on the timeline and in each clip's own transcript paragraph. Material that plays once behaves as always.
• Through edits: a cut between two clips where nothing is missing is drawn as a dashed line. Right-click either clip and choose Join with Next Clip or Join with Previous Clip to remove the cut.
• Angle labels: in a multicam project every clip carries a small label with the angle it shows, so you can read the switching at a glance.
• Cuts made by hand show in the transcript. Words removed by deleting or trimming on the timeline, or with the Range tool, read as cut with a red line through them. Right-click one and choose Restore This Cut to bring the material back.
• A plain click always selects a clip, even with the Range tool on; the Range tool only changes what a drag does.
• In the full-space view every kept section between cuts has its own grips, so you can widen or tighten one keep without touching its neighbours; an outward trim stops at the next clip.
• Fewer edits for nothing: a pause cut inside a take must remove at least a quarter second, tiny stubs left between two cuts fold into the cut, and breath cuts respect the pause slider.
• Right-click any tinted cut to restore it. ⌘= and ⌘− zoom the timeline around the playhead (with Follow on) or the selected clip.

Zoom & punch-ins

Punch in on the moments that deserve it, straight from the paper edit, and export the zooms as real transform adjustments instead of doing keyframe work in the NLE.

• From the transcript: select the words to zoom on and pick a level from the selection bar's zoom menu. Two presets ship at 110% and 125%; both are editable in the Zoom panel on the right.
• From the timeline: right-click a clip and choose a zoom level for that whole piece. Right-click again to remove it.
• Aim it in the preview. While the playhead sits inside a zoom, the preview shows the punch-in live. Drag the dot to aim the zoom at a face or a detail, and pull the ring around the dot to fine-tune the strength, with the percentage shown as you drag.
• Zoomed spans are marked in the transcript with an indigo underline and on the timeline as indigo strips. Zooms survive re-transcription and ride along in shared projects.
• Exports are frame accurate and resolution independent, at any project size including 4K. Final Cut Pro and DaVinci Resolve receive each zoom as a transform adjustment on the cut, on plain clips, multicam and compound clips alike. Wherever a zoom's edge lines up with an existing edit, PowerCut reuses that edit point instead of adding a new cut. Premiere Pro receives the cuts without the transform, since its XML format has no equivalent that survives the import.

Markers, keywords & favorites

Mark up the paper edit the way Final Cut thinks, right in the transcript or the Document view. Select words (or just click one) and press F to favorite, K to add a keyword, or M to drop a named marker. Right-clicking any word offers the same three, plus removal of whatever already sits there. Favorites underline green, keywords amber, and markers show as teal bookmarks above their word. The timeline shows all three too: green and amber strips along the top of clips, bookmark ticks for markers.

Working with markers. Click a bookmark to select it and open its editor: rename, save, or delete. Right-click for the same, or press Delete with one selected. Esc or a click on empty space deselects.

The Markers & Tags list. The bookmark button in the transcript header opens a list of everything in the project. Click a row to jump to that spot in the transcript and timeline, rename markers inline, or delete anything. The eye chips at the top hide any of the three kinds everywhere in the app (exports always include everything), and the Edit menu can hide kinds or remove all of a kind in one undoable step.

Export list. The Export List button (also File → Export Markers & Tags) saves the whole set as a CSV, or a plain text document if you change the extension to .txt. Each entry carries its type, name, clip, source timecode as the camera wrote it, timeline timecode matching the XML export, and the sentence it sits on.

How they export. Final Cut imports keywords and favorites onto the original clips in the browser, and markers onto the timeline, the source clips, and inside multicams. Markers on struck words stay out of the edit but still ride inside the multicam. DaVinci Resolve receives markers by EDL sidecar (see Exporting). Premiere receives timeline markers.

Working with a script

If the video was read from a script, load it in the Script section of the Edit tab: paste it, import the file (Word .docx, RTF, Markdown or plain text), or simply drop the document anywhere in the PowerCut window and the Script sheet opens with it. Pages documents need a quick File, Export To, Word first. PowerCut then:

• corrects branded spellings in the transcript ("Ben Q design view" becomes "BenQ DesignVue"), applied instantly, no re-transcription;
• picks takes by how well each read covers its script line;
• strikes off-script asides, verified two ways so reordered reads are never harmed.

The script is guidance, not a filter. When the chosen reads cover less than half of what the clip says (a talk given from notes, a script that was only a starting point), PowerCut treats the script as loose: it still picks the take of a repeated line and still corrects spellings, but it strikes no off-script words and no stumbles, and the Script section says so. Removing a script offers to run the Auto Edit Pass again without it.

Adding a script after transcription offers to re-run the pass with the new guidance. Every automatic spelling correction shows its work: corrected words carry a dotted teal underline, hovering shows what was actually heard, and right-click offers Revert Correction. The Script section counts the corrections it has made.

Finding the story EXPERIMENTAL

Find the Story in the top bar opens the Footage Brief, a panel that reads the whole shoot (not just the cut) and tells you what you actually have: the beats in the order you filmed them, the strong themes and how many spoken seconds cover each, candidate openings and endings, and watch-outs like thin coverage or beats that wander off the thread.

Said more than once. Ideas you delivered several times are grouped, with a verdict on which read is strongest: more energy, fewer fillers, or the one that finishes the thought. Each row can play the read or show it in the transcript, and tells you whether the strongest read is the one currently in your cut. A HIGH ENERGY tag marks reads clearly above your own typical energy in that shoot.

Labels in the transcript. Each paragraph can carry its story role (hook, setup, problem, result, reflection and so on) as a small chip and a colored rail; toggle them from the top bar. Clicking an item in the brief highlights the beat in the transcript in its role color; click anywhere else and the highlight clears.

Everything here is computed on your Mac; with Apple Intelligence enabled the labels get a second, smarter pass. It is experimental: treat the brief as a well-prepared assistant's notes, not as instructions.

Assistant Editor + BETA

Completely optional. PowerCut works fully without this. Assistant Editor + lets the AI app you already use, the Claude Desktop app, Codex or Cursor, work on your project with you: you chat with it, it reads the transcript and makes the changes in PowerCut, and you watch them land, each with the assistant's reason and an Undo. No key to create and nothing to pay PowerCut; the AI runs on your own account.

How it works. PowerCut speaks MCP, the open protocol AI apps use to reach tools on your Mac. Once connected, the AI app sees PowerCut's tools: read the transcript, search it, list the take groups and the review queue, and then strike, restore, correct a misheard word, pick a take, settle a review item, add markers, keywords and favorites, move or split paragraphs, change a cleanup switch, or run the Auto Edit Pass. Every change requires the assistant to give a reason and lands as one undoable step named after it ("Claude: Strike"). Reads are free of side effects.

Connecting. Open PowerCut's Settings and choose the Assistant tab:

• Claude Desktop: click Connect Claude Desktop. Claude Desktop rewrites its settings file while it runs, so PowerCut asks to close it, adds itself to the configuration (the previous file is backed up beside it), and reopens it. A tools icon under the message box then lists PowerCut's tools, and you can ask it anything about the project you have open in PowerCut.
• Codex: click Connect Codex. PowerCut adds itself to Codex's config.toml; a new Codex session lists the tools. If Codex already has a PowerCut entry, PowerCut shows the snippet to paste instead of rewriting your file.
• Cursor: click Connect Cursor. PowerCut adds itself to Cursor's global ~/.cursor/mcp.json (the previous file is backed up beside it; a project's own .cursor/mcp.json is never touched). In Cursor, open Settings, MCP and switch PowerCut on if it is off; restart Cursor if it does not appear. The transcript then goes to whichever model provider you chose in Cursor.
• Other MCP apps: Copy command puts the path of PowerCut's helper on the clipboard; add it as a local (stdio) server in that app. ChatGPT's desktop app only takes remote connectors and cannot reach PowerCut on your Mac yet.

PowerCut has to be open with a project loaded. If the AI app calls while PowerCut is closed, the helper opens PowerCut and waits for it. The switch Let AI apps connect to PowerCut turns the connection off entirely.

What you see. The Assistant tab under the preview shows every read and every change, newest first: "Claude struck 3 words in 1 sentence: …" with the reason beneath, and an Undo button while that change is still the latest. The transcript, timeline and review queue update as the assistant works, so nothing happens out of sight. Three starter prompts sit in the tab with a copy button.

Prompts that work well. "Read my transcript and tell me where the pacing drags." "Review the take picks and swap the ones you are sure about, with a reason for each." "Strike anything that repeats an earlier point, one sentence at a time." "Select the part where I explain pricing." Ask it to explain before it acts when you want to stay in control; ask it to go ahead when you do not.

Safety. Writes are refused while a clip is transcribing or the Auto Edit Pass runs, and refused when the project changed since the assistant last read it (it has to read again). The assistant can undo its own latest change; anything older is yours to undo. Versions are unaffected: the Auto Edit Pass still saves one first, and the assistant can save one on request.

Not on-device. This is the one feature that sends anything off your Mac, and PowerCut says so in Settings and asks you to confirm before it connects an app. When the connected AI asks, PowerCut hands it the transcript of the open project (every word with its timing, pauses and speaker names), your script, the clip and project names, take groups, the review queue, your annotations, your selection, the cut plan, and the reason for each change it makes. That text travels through the AI app to its provider, Anthropic for Claude or OpenAI for Codex, on your own account and under their terms. Audio, video and your files never leave your Mac, PowerCut sends nothing on its own, and the switch Let AI apps connect to PowerCut turns it all off. Learn More in the confirmation walks through what is sent, how, and why.

Versions

The Versions button in the transcript header lists timestamped copies of the whole edit. One is saved automatically before every Auto Edit Pass; save one manually any time (⌥⌘S). Two ways back:

• Revert replaces the current edit with the version (undoable, like everything).
• Merge lists every place the version and the current edit disagree about what's in the cut, quoted line by line. Adopt the version's side or keep yours, per difference, with a jump-to for judging in context.

Sharing projects

A .powercuts file is a plain document with no license lock: anyone with PowerCut can open it, and the free version is the full editor (a license matters only for export). The transcript travels INSIDE the file (every word, strike, correction, and per-clip language) so the recipient never re-transcribes, and a project made on macOS 26 opens fine on macOS 14.

Opening a project. Double-click a .powercuts file in the Finder and it opens in PowerCut (or drop it on the app icon, or use File → Open).

Media travels separately. Clips are referenced, never copied. The simple recipe: put the media in the same folder as the .powercuts file (or one folder inside it), zip the folder, send it. On open, PowerCut finds the files next to the project automatically; every automatic match is verified against the clip's stored metadata, so a same-named but different file is refused rather than silently taking a clip's place.

Save Project with Audio. When the media is too big to send, or the recipient only needs to hear the edit, choose File → Save Project with Audio… It writes one folder: the project file plus a small AAC copy of every online clip's dialogue, exactly as PowerCut hears it (the chosen track, channel or mix). Zip the folder and send it, or keep it as an archive. Opened from that folder, the project has the transcript, every cut and audible playback; the clips show no picture, angles are not included, and an export references the audio copies, so someone with the footage relinks first. Offline clips and clips without audio keep their original paths. Your own project is untouched: the folder is a copy.

Relink Media. When clips are offline (moved drives, renamed folders), choose File → Relink Media… or click the amber banner in the sidebar. The dialog lists every offline clip with its own Locate… button (external drives and scattered folders welcome) plus Match a Folder… to point the whole project at one place. Right-clicking a clip in the sidebar still relinks it directly.

Missing media is not a crisis. Offline clips keep their entire edit and keep working from saved data: read, search, strike, restore, and export XML referencing the original paths. Only playback and re-transcription need the file. Two things that do NOT travel with the file: the Versions history (it lives in a folder beside the project) and learned corrections (they belong to the app). And since the transcript is readable inside the file, share accordingly.

Multicam angles

Give each clip the other cameras and recorders that rolled alongside it, and the export arrives in Final Cut Pro or DaVinci Resolve as one synced multicam, ready to angle-switch. Choose Multicam For Each Clip in the Export sheet and open Angles… next to it (the all-clips multicam holds every source in a single angle and has no extra angles). Add the files that belong to each clip, name every angle the way you want it to appear, and export. You can also create named blank angles now and choose their clips later; blanks arrive as empty lanes you can fill inside the NLE. Each angle with audio also has a Dialogue checkbox: tick it and that angle's audio is mixed into the primary clip's dialogue at its offset, so transcription, analysis and playback hear both; a Mic menu picks which of the angle's tracks carries the voice. The checkbox stays off until the angle's sync is known (a manual offset, Sync by audio, or matching timecode on both files), because a wrong offset would smear two speakers. Angles are remembered with the project and media is referenced in place, never copied. Create Multicam Clip (File menu, or the empty project screen) builds the set in one step from a folder of files, and its main clip can be audio-only: a field recorder's WAV with the cameras as angles. Every file with audio lists its tracks with three boxes: Transcribe (the tracks the transcript reads; the main clip may tick several, an angle one), Hear (heard while you edit, never transcribed) and Sync (the one track the audio match listens to), so the transcript reads the mic track while the match listens to whichever track suits it. Each angle shows how it synced, and Check Sync plays a few seconds with the main clip in your left ear and the angle in your right: matching timing sounds like one voice, an offset like an echo. On the New Project page, double-click a multicam to open it again and change its angles, names, tracks or sync; only a change to the audio the transcript reads transcribes again. Multichannel recorder files transcribe like any other.

All the files at once. Drop every file of the shoot on the Create Multicam sheet in one go, or pick several in the Choose dialog: each becomes an angle named after its file, rows are added as needed, and the first camera file becomes the main clip when none is chosen yet (recorder files such as m4a or WAV never do). Dropping on a particular row fills that row first, then the empty rows below it. Rename any angle afterwards; the name travels into the NLE.

Angle order. Drag the grip at the left of a row to move it: the row lifts under the cursor and the others slide aside. The chevrons on the right do the same one step at a time. The order is the angle order in the Final Cut Pro multicam and the track order in Resolve and Premiere. Rows still called "Angle 1, 2, 3" renumber when they move; names you typed stay.

Switching angles. Select words in the transcript and choose Picture in the right-click menu, or Angle in the floating bar above the selection, to show another angle's picture for those words. The switch applies wherever that material appears in the story, the player shows it, and the export writes it as the multicam's picture for that stretch (Multicam For Each Clip; in Premiere the switched angle's clip arrives enabled on its stacked track; other wrappers ignore switches). Angles that cannot show say why; an angle that only covers part of the selection shows the main clip for the rest. Importing a multicam brings its picture switches in the same way (audio switches later in the timeline are not carried). Remove a switch from the same menu. Automatic switching by speaker is next.

Speakers. Labels are opt-in: tick Multiple speakers on the empty project screen or in Create Multicam Clip (off by default, since a one-person clip needs no names), or turn them on later from a clip's menu in the Footage panel. When a clip's dialogue comes from separate microphones, one track or angle per person, PowerCut listens to each microphone on its own and labels who is talking: a colored name in front of the first word of each turn in the transcript, a name under each paragraph's timestamp in the Document view, and, if you tick it in the captions export, a "Name: " at the start of every caption (a change of speaker starts a new caption). Speakers are named after their angle or track; rename them, merge two that are the same person, and give each the angle that should show when they talk (for automatic switching, coming next) in the Speakers section of the Edit tab, where Hide Speaker Labels takes the names off the transcript without discarding them. Words two people say at once, and silence, carry no name. One microphone with several people on it works too: PowerCut identifies the voices themselves, downloading a small model once (about 21 MB); set how many people talk in the Speakers panel if it guesses wrong, or leave it on auto. Detect Again re-listens after you change the mics.

The main angle. The player's angle menu, next to the transport buttons, picks the main picture for the source under the playhead, everywhere that source appears: the player shows it and the Final Cut or Resolve multicam export uses it as the picture, with words switched in the transcript keeping their own angle. The audio stays the dialogue mix, an angle that is offline or not yet synced says so and keeps showing the main clip, and any stretch the angle does not cover shows the main clip too. Rename an angle from the same menu (or from the transcript's Picture submenu, or in Angles); the name travels into the NLE.

Export folder. Exports open next to the footage the first time; pick another folder and that project remembers it for every later export until you pick a different one.

Three ways to sync, per angle, audio first. Sync by audio is the default: as soon as an angle's file is read, PowerCut listens to the clip's selected audio track (the Footage panel's Track choice) against the angle's track and matches the waveforms, no timecode needed. When the waveforms cannot be matched and both files carry usable timecode (a real timecode track, or the Broadcast Wave timecode a field recorder writes into its WAV files), timecode takes over and the angle says so; Sony metadata timecode syncs in Resolve only. With neither, the angle gets a manual offset of 0 s, an aligned start you can adjust, so a multicam is never held up by a match that cannot happen. Timecode and a manual offset are also choices in the angle's menu. If an angle has several audio tracks, pick the one to listen to; a file with no audio says so.

Exporting

Click Export in the toolbar: the Export sheet asks for the destination (Final Cut Pro, DaVinci Resolve or Premiere Pro), holds the options and the transcript and caption exports, and Export XML writes the timeline. The export is a cuts-only, frame-accurate timeline that references your original media untouched.

Resolve and clips with several audio tracks. DaVinci Resolve's XML import brings in a clip's first audio track only, in both XML formats: a long-standing Resolve limitation. PowerCut handles it two ways. A plain timeline for Resolve that contains a clip with more than one audio track (a screen recorder's mic and system audio, a camera with a second mic) is written as OpenTimelineIO instead: a .otio file Resolve imports the same way (File → Import → Timeline), one audio track per source track, every track playing. If your export wrapper is a compound clip or a multicam, PowerCut asks whether you want plain clips as .otio or the wrapper as XML. With the wrapper, fix the audio once per source after the import: open the compound clip or multicam from the media pool, select the clip inside, right-click, Clip Attributes, Audio, and set the Source Channels in increasing order (Embedded Channel 1, 2, 3 and so on). Every use of it on the timeline follows. Final Cut Pro reads all tracks from the XML directly.

Wrappers (Final Cut Pro and Resolve). The wrapper menu in the Export sheet decides how your sources arrive. Plain Clips: each kept piece is a clip straight off the source, the safest handoff. Compound Clips For Each Clip: every source is wrapped in its own compound clip, and the kept pieces are slices of it, so you can open a compound and extend any piece with the rest of that take. Compound Clip All Clips: one compound clip holds every source back to back, in story order, and the whole edit is slices of that single clip; handy when you want one thing to open, retime or re-cut. Multicam For Each Clip: every source becomes an angle-switchable multicam, with room for the extra cameras and recorders you attach under Angles. Multicam All Clips: one multicam with every source back to back in a single angle. Markers on the original footage ride inside every wrapper, and punch-ins export as real transforms in all five modes.

Vertical projects. The exported sequence follows the footage: phone or vertical clips produce a vertical (1080 by 1920, or 4K vertical) sequence in every NLE; landscape footage keeps the broadcast sizes as before. Mixed shoots follow whichever orientation carries most of the kept material.

Where the file goes. Every export dialog opens in the folder of your first clip, so the XML lands next to the footage by default; change the folder in the dialog if you keep exports elsewhere.

The export dialog. The Transcript and Captions buttons (and File menu, and the Document view's Export) open one dialog: pick what to include, then the language: the original languages, or English · translated. The translation happens at export time only: the project and your edit are never changed, strikes map onto the English words by time, and translated clips are cached so repeat exports are instant. Files are tagged (English). So a Hindi edit ships with English captions or an English producer script in one step. Clips still missing a transcript are called out in the dialog rather than silently left out.

Camera timecode. Sony MP4s (FX3, FX6, Alpha bodies) carry their start timecode in camera metadata that macOS does not surface. PowerCut reads it straight from the file and writes it into the export, so DaVinci Resolve links the clips instead of reporting them missing, and your NLE shows real camera timecode.

Keyboard shortcuts

This list is one click away inside the app: Help, Keyboard Shortcuts.

License & free version

Updates & ownership. Your license is permanent. Every update through version 1.x is free, including macOS compatibility fixes. A future major version will be a paid upgrade, always discounted for existing owners. If this product is ever retired I will say so rather than go quiet: the final version stays downloadable and your license keeps working.

PowerCut is a one-time purchase, no subscription. New installs get 7 days fully functional, export included. After that the app becomes the free version: the entire editor keeps working forever, projects keep autosaving, and only timeline and caption export lock. A license key (delivered by Payhip after purchase) unlocks export instantly, including for everything you edited while free. Enter it under the app menu → Enter License Key (⇧⌘L).

One license, two Macs. A license is personal and covers the Macs you use yourself, up to two at a time, for example a desktop and a laptop. Enter the same key on both. When you replace a Mac, choose Remove License from This Mac in Manage License on the old one first, then activate the new one. If the old Mac is gone, entering the key on a third Mac offers Release Both and Use Here, which frees both seats at once; a released Mac keeps working until its next license check, then drops to the free version. A second editor needs a second license.

How activation works, and what leaves your Mac. When you enter a key, PowerCut sends it, together with a random id for this Mac, to a small license server Rafael runs on Cloudflare. The server checks the key with Payhip and answers with a signed license that PowerCut keeps in your Keychain. The server stores only a hash of the key and of your purchase email, the plan, purchase dates and the two Mac ids. Nothing about your projects, footage or transcripts is ever sent. PowerCut re-checks the license about every two weeks when online; if the server cannot be reached, everything keeps working.

Lifetime and 18 Months. A Lifetime license includes every update. An 18-Month license includes every update released within 18 months of purchase; after that PowerCut keeps working forever at the last version you received, and you can renew for another year or move to Lifetime at any time. Settings shows which you have.

Updates arrive through the built-in updater (app menu → Check for Updates), signed and verified.

Beta builds. New features ship as a beta first. PowerCut Beta is a separate download from the beta page (also reachable from Settings, Updates, Try the Beta). It installs beside PowerCut, opens the same projects, updates itself, and your license works in both; without one it has its own 7-day trial. The beta never writes a release project: opening "Interview.powercuts" makes "Interview Beta.powercuts" next to it and works on that copy, so the original is untouched. To carry a cut you made in the beta into PowerCut, open the Beta copy there. The beta page explains how to send feedback and what to expect.

Uninstalling

PowerCut is a single, self-contained app. To uninstall, quit PowerCut and drag PowerCut.app from your Applications folder to the Trash. That's the whole app gone.

Optional: remove support files

A few small files remain if you want a completely clean sweep. In Finder, choose Go → Go to Folder and remove these:

• ~/Library/Preferences/com.rafaelludwig.PowerCuts.plist (settings)
• ~/Library/Saved Application State/com.rafaelludwig.PowerCuts.savedState (window state)
• ~/Library/Application Support/PowerCuts (internal autosave from early versions)
• ~/Library/Application Support/FluidAudio (the verbatim engine's ~500 MB speech model; shared cache, safe to delete)

Your license is not lost by uninstalling: the key from your Payhip receipt activates any future install. If you're leaving because something didn't work, one email might fix it: apps@rafaelludwig.com.

Troubleshooting

• "On-device transcription needs macOS 26." Transcription requires Apple's speech engine on macOS 26 (Tahoe) or newer.
• "Verbatim engine unavailable." The second engine could not process a clip (usually its one-time model download was interrupted, or an unusual container). Filler and stutter detection are skipped for that clip; everything else works. Toggle "Remove filler words" off and on to retry.
• A take pick or strike looks wrong. Check the Review queue first; then hover badges for reasons, click to swap, restore with U. Every override teaches the rules.
• The app froze. If PowerCut stops responding for more than a few seconds, it records what it was doing at that moment in ~/Library/Logs/PowerCut/hangs.log. Force quit, open PowerCut again and choose Help → Report a Problem: the report is attached, and it is what makes the freeze fixable.
• The app crashed. PowerCut offers to send the crash report on next launch (you review the email before anything is sent), or Help → Send Crash Reports.
• Something else. Help → Report a Problem pre-fills an email with your version details. Suggestions are just as welcome: Help → Leave a Suggestion.