For Claude video editing, start by deciding whether you need written edit notes, a code-based animation, or a timeline you can keep adjusting. To learn how to edit video with Claude on a live multi-track timeline, connect Claude Code—either inside the Code tab of the Claude desktop app or the claude CLI—to ChatCut Desktop on macOS 13+ or Windows 10+ (x64; Linux is not supported). We make ChatCut, the AI video editor connected to Claude Code in this guide.
Below, we walk through turning one unedited 2-minute 48-second (168.2-second) attic talking-head recording into two finished 60.0-second 1080p deliverables—a 16:9 horizontal edit and a 9:16 vertical Short—complete with filler and pause cleanup, active-word orange captions, spring punch-in zooms, animated Motion Graphics, a pre-quoted AI B-roll cutaway, and auto-ducked background music. In this shared ChatCut Desktop master project, the original eight-step elapsed window was 13 minutes 46 seconds, excluding later B-roll retries. The 4-second Seedance 2.5 cutaway used in the finished edit cost 4.38 ChatCut credits; removing garbled screen text required three generations in total (13.14 credits), including the original attempt and two retries. Every cut, caption, graphic, audio mix, and local export cost 0 ChatCut credits.
The chat windows below are code-built illustrations of Claude Code, not live Claude screenshots. The editing itself was a real shared master project in ChatCut Desktop, driven by an external agent (Codex). Sending the same prompts from a connected Claude Code session gives it access to the same ChatCut tools; edit decisions, timing, and results can differ. All timings here belong to the original master run; generation charges include its later B-roll retries. These are not a Claude benchmark. For a separate discussion, see our guide to Claude video editing.
How does Claude video editing work in Claude Code and ChatCut?
Which Claude surface you open determines whether you get back text notes, a code animation, a flat rendered file, or an editable timeline project:
| Claude route | Where it runs | What it actually does | What you get back |
|---|---|---|---|
Regular Claude chat (claude.ai or mobile apps) | Web browser, iOS, Android, or Desktop Chat tab | Reads transcripts or scripts you paste; Anthropic’s file upload guide lists documents and still images, but does not list video uploads | Transcript-based edit notes; this route does not connect to the local ChatCut timeline |
| Claude Motion | Beta on Claude Team and Enterprise plans only | Generates short code-based animations from text, data, and images | Shareable Motion artifacts and MP4 exports for charts or explainers, not cuts of camera footage |
Claude Code + terminal toolkits (e.g., video-use) | A local agent session with the toolkit dependencies installed | Uses scripts and FFmpeg for cuts, captions, and animation overlays | A rendered MP4 plus project files; revised exports need rendering again |
| Claude Code + ChatCut plugin | Local Code tab session, claude CLI, or inside ChatCut Desktop (macOS 13+ / Windows 10+ x64) | Drives ChatCut Desktop’s timeline, transcript, caption, Motion Graphics, and AI generation tools over MCP | An editable multi-track timeline you can tweak by hand, plus local MP4/4K or XML handoff exports |
If you are comparing third-party open-source repositories across analysis, code animation, and timeline editing, see our breakdown of Claude Code video editing skills and tools and our broader roundup of the best Claude Code plugins. If your goal is purely animated typography, charts, and lower thirds, see our Claude motion graphics walkthrough.
Where Claude Code connects, which models it runs, and how billing splits
- Run Claude Code locally on the same Mac or Windows PC as ChatCut Desktop. In Anthropic’s Claude desktop app, select the Code tab and set the environment selector below the prompt box to Local (a Cloud or SSH session runs on another machine and cannot reach ChatCut Desktop on yours). You can also run
claudein Terminal or PowerShell (Claude Code overview), or stay inside ChatCut Desktop and pick Claude Code under Agent Selection at the bottom of the AI composer (ChatCut Desktop guide). The desktop Code tab and CLI support personal skills in~/.claude/skills/; configuration and tool availability also depend on the session and project. Need to continue an existing cloud project inapp.chatcut.ioinstead? Install the separate hosted Web plugin, which opens ChatCut Web in the desktop app’s Browser pane. - Your own Claude models and subscription cover the agent. Claude Code runs on the models configured on your Anthropic account—under Claude Code’s model configuration, the standard
opus,sonnet, andhaikualiases resolve on the Anthropic API to Claude Opus 5.5, Sonnet 5.5, and Haiku 5.5, andopusplanplans complex multi-step jobs on Opus before executing on Sonnet. Individual access is included with Claude Pro, Max, and Ultra plans; organization access depends on the assigned Team or Enterprise seat. Anthropic Console API usage is billed separately. (Tip: if you use a monthly Pro or Max subscription, make sureANTHROPIC_API_KEYis not exported in your shell environment, or Claude Code will default to pay-as-you-go API billing.) - Core timeline editing costs 0 ChatCut credits. When Claude Code drives ChatCut over MCP, Claude’s reasoning uses your Claude plan rather than ChatCut’s built-in Agent meter (Credits policy). Importing local clips, cloud speech transcription, rough cuts, captions, library sound effects, agent-authored Motion Graphics, and watermark-free local MP4 exports cost 0 ChatCut credits on ChatCut’s Free plan (Free caps cloud-rendered exports at 60 cumulative minutes, while ChatCut Desktop renders video locally on your computer). You spend ChatCut credits only when you approve ChatCut-hosted AI media generation; AI video and image generation require a paid ChatCut plan (starting at $25/month for 100 credits).
What did our 8-step single-take edit produce?
Here is a 1280×720, 30 fps web preview (6.02 MB) of the finished 60-second 16:9 edit. The shared master was exported at 1920×1080 (94.60 MB); this smaller preview preserves the edit timing but resizes and recompresses the picture. The source recording was 168.2 seconds long:
Watch five distinct layers come together across those 60 seconds:
- 0:00 to 0:03: A transparent animated opening title (
CHATCUT · THE NEW RELEASEabove “Make ideas move.”) with an expanding warm orange accent rule. - 0:04 to 0:06: Bold white lower-third captions (
Arial 800) where only the currently spoken word turns orange (#FF9B40) and returns to white once spoken, with the brand name checked in the caption layer. - 0:07.5 to 0:13.5: A staggered three-point Motion Graphics callout (
01 Stability + performance,02 A redesigned UI,03 AI Motion Graphics) paired with the first 1.18× spring punch-in zoom from 0:09.3 to 0:12.7. - 0:36.4 to 0:46.5: A subtle library whoosh (-14 dB) marking the third point, a second 1.18× spring zoom (0:38.5–0:42.4), and a 4-second Seedance 2.5 cutaway slowly pushing toward a monitor on a wooden desk in a warm-lit attic, with abstract orange and colored blocks on screen (0:42.5–0:46.5, 4.38 credits for the final clip; 13.14 credits across three generations to remove garbled screen text) while the speaker’s voice and captions continue underneath.
- 0:00 to 1:00: A free Creative Commons (CC BY 4.0) ambient instrumental bed faded in over 1.5 seconds, automatically ducked beneath the voice (-20.3 LUFS integrated mix), and faded out over 2.0 seconds.
How do you edit a video with Claude in five steps?
- Install and connect ChatCut once. Open a Local session in the Code tab of the Claude desktop app (or run
claudein your terminal) on macOS 13+ or Windows 10+ (x64), send the one-line setup prompt, sign in inside ChatCut Desktop, and open a New session so Claude loads ChatCut’s local editing tools. - Import your raw take and cut the A-Roll. Point Claude at your recording, let cloud transcription finish, and ask for a tight spoken cut that removes fillers, false starts, repeated takes, and dead air while preserving your core argument.
- Add active-word captions, zooms, and subtle sound cues. Style lower-third captions, verify brand spellings, and place restrained punch-in zooms and library sound effects on real topic shifts.
- Layer Motion Graphics, pre-quoted B-Roll, and ducked music. Add animated titles and structured point callouts (0 credits when authored as code layers), require a credit quote before generating any AI cutaway footage, and lay in an auto-ducked music bed.
- Export horizontal and vertical deliverables—and inspect the files. Render a local 1080p (or 4K) 16:9 MP4, duplicate the cut onto a 9:16 Shorts timeline with adjusted text and crop framing, and play both MP4s end to end before publishing.
What do you need before you start?
- A Mac (macOS 13+ on Apple silicon or Intel) or Windows PC (Windows 10+ on x64). ChatCut Desktop supports macOS and Windows only; Linux is not supported (ChatCut Product Facts). You also cannot install ChatCut on
claude.aiin a browser or in the Claude iOS/Android apps. - Claude Code with an active Claude plan or API billing. Use the Code tab in the Claude desktop app (with a Local environment selected) or
claudein your terminal, signed into an eligible Claude subscription or Anthropic Console account. Organization users should check that their assigned seat includes Claude Code. - A signed-in ChatCut account. Sign in inside ChatCut Desktop so cloud speech transcription and built-in sound libraries are available. Free covers timeline editing, captions, Motion Graphics, and local exports; add a paid plan only if you plan to run ChatCut-hosted AI video or image generation.
- One clear source clip and a credit ceiling. Keep your raw
.mp4or.movin a local project folder—optionally alongside ascript.mdoutline orCLAUDE.mdstyle checklist—and state your maximum credit budget (such as 5 credits max, quote first) before asking for generated B-roll.
Why should you lock A-Roll before adding B-Roll?
For this walkthrough, organize the edit into two stages:
- A-Roll is the spoken backbone of your video: the camera recording, transcript cuts, word-level captions, punch-in zooms, and transition sound cues (Steps 1 to 4 below).
- B-Roll is everything stacked above or below that backbone to clarify or score it: animated title cards, numbered callouts, AI cutaway shots, and background music (Steps 5 to 7 below).
Set the spoken timing first, then place overlays and music against it. Later trims may require checking those layers again. In the shared master run, Steps 1–4 occupied about 3 minutes 44 seconds, including the short gaps between steps; this is an observation from one run, not an expected Claude completion time.
How do you edit video with Claude step by step?
First connect Claude Code to ChatCut Desktop once; then run Steps 1 to 4 to finish the A-Roll and Steps 5 to 8 to add B-Roll and export both aspect ratios.
Setup: connect ChatCut to Claude Code in one prompt
Open the Claude desktop app, switch to the Code tab, confirm the environment selector under the prompt box is set to Local, and paste the setup prompt from the ChatCut Claude Code plugin page (also used across chatcut.io/desktop):
You can also use the Claude-specific /goal setup prompt documented in ChatCut Product Facts, which fetches chatcut.io/claude and runs the same native Desktop installer:
(If you are continuing an existing cloud project in app.chatcut.io instead of editing local files, send /goal Read chatcut.io/web-plugin/claude to install and use the ChatCut plugin from the Web plugin guide.)
Claude reads the Desktop installation instructions and runs ChatCut’s official installer (curl -fsSL https://chatcut.io/install.sh | sh on macOS or irm https://chatcut.io/install.ps1 | iex in Windows PowerShell—no Node.js or npm required). Once ChatCut Desktop opens, complete sign-in in your browser. Desktop automatically registers its local MCP server with Claude Code and copies ChatCut’s editing workflows into ~/.claude/skills/ without overwriting skills you wrote yourself.
For this connection, follow the ChatCut installation guide: click + New session in the Code tab (or exit and relaunch claude in your terminal) after signing in. Claude Code also provides MCP connection controls. Verify this connection with a read-only check before touching footage:
Step 1. Import the raw attic recording and transcribe speech
Our source clip (mg-teaser.mp4) is a single 2-minute 48-second (168.167-second, 5,045-frame, 1920×1080 at 30 fps) take of a presenter in an orange shirt speaking inside a sunlit attic room, complete with false starts, repeated explanations, and pauses between thoughts. Start with import and speech transcription only:
You: Import this clip into ChatCut and transcribe it.
Claude: Imported mg-teaser.mp4 (2:48, 1920×1080 at 30 fps) onto the timeline and completed word-timed speech transcription. Ready for the rough cut.
In the shared master run, importing the local file and completing word-timed cloud transcription took 38 seconds (0:38) and 0 ChatCut credits. Because ChatCut Desktop reads local video files directly from disk (Desktop media import), editing and local export can use the source file on your computer. Speech transcription still uses a cloud service.
Step 2. Cut 2 minutes 48 seconds down to a clean 60-second A-Roll spine
With the word-timed transcript ready, ask for a 60-second 16:9 rough cut that strips out verbal clutter while preserving the core three-part release announcement:
You: Make a clean 60-second 16:9 cut: remove filler words, false starts, repeated lines and long pauses. Keep the main argument.
Claude: Trimmed the 168.2s recording to an exact 60.0s (1,800 frames at 30 fps) across 8 clean segments—keeping the hook, stability & performance, the redesigned UI, AI Motion Graphics, and the sign-off while cutting 7 filler words, retakes, and three multi-second pauses.
In the master run, text-based editing handled the trim in two passes inside 1 minute 33 seconds (0 credits):
- Mechanical filler cleanup removed 7 spoken fillers (4 “uh,”, 2 “um,”, and 1 “ah,”, trimming 2.07 seconds immediately).
- Semantic script selection + pause tightening kept the opening hook, all three product points, and the closing line while dropping retakes and tangential side examples (66.43 seconds), then tightened three remaining gaps (1.60s → 0.35s, 2.16s → 0.40s, and 4.40s → 0.97s) to hit exact 60.0 seconds (1,800 frames at 30 fps) across 8 audio-video cuts.

Open ChatCut Desktop and play across all clip joins before moving on. Check that speech remains natural at every join.
Step 3. Add lower-third captions that highlight only the active word
Once the spoken timing is locked, add AI captions styled to match the warm palette of the shot:
You: Add word-by-word captions: bold white text, the spoken word highlighted in orange, placed in the lower third.
Claude: Added bold white lower-third captions (Arial 800, 64 px) with an editable HTML caption template that highlights only the active word in orange (#FF9B40), and checked the brand spelling in the caption layer.
In the master run, 1 minute 6 seconds (0 credits) covered building an editable HTML caption template driven by each word’s exact start and end timestamps: the word being spoken turns warm orange (#FF9B40), already-spoken words return to white. Caption text was corrected separately from the audio. The W3C WAI captions guide discusses caption positioning and styling. In this edit, the lower-third placement keeps the speaker’s face visible; check placement against your own footage.

Step 4. Finish the A-Roll with two spring punch-in zooms and a subtle whoosh
Before adding graphics, ask for two restrained zooms and a sound cue at a topic change:
You: Add two punch-in zooms on the key lines and a subtle whoosh where the topic changes.
Claude: Added 1.18× spring punch-in zooms at 0:09.3–0:12.7 and 0:38.5–0:42.4, plus an Airy Short Whoosh from ChatCut's sound library at 0:36.4 (-14 dB) where the third topic starts.
In the master run, that pass finished in 25 seconds at 0 ChatCut credits: two 1.18× spring zooms (9.33–12.67 s on stability & performance and 38.50–42.37 s on AI Motion Graphics) add visual emphasis without cropping the speaker’s forehead, and the built-in Airy Short Whoosh (36.43–37.20 s at -14 dB) marks the transition into the third section without spending generation credits. That locks the A-Roll.
Step 5. Add an animated opening title and a 3-point Motion Graphics callout
Now the B-Roll pass begins. Rather than flattening static text onto the video, ChatCut places editable AI Motion Graphics on their own timeline tracks:
You: Add a 3-second animated title at the start and a Motion Graphics callout that lists the three points I make.
Claude: Authored two transparent HTML Motion Graphics layers in matching warm editorial type: 'Make ideas move.' (0:00–0:03) and a staggered 'THREE IMPROVEMENTS' callout on the right side of the frame (0:07.5–0:13.5).
In the master run, 1 minute 51 seconds (0 ChatCut credits) covered this step: two transparent HTML Motion Graphics components landed above the footage:
- 0:00 to 0:03 (
Make ideas move.): Centered editorial headline beneath aCHATCUT · THE NEW RELEASEeyebrow, entering with a smooth upward fade and an expanding orange divider rule. - 0:07.5 to 0:13.5 (
THREE IMPROVEMENTS): Right-aligned numbered list (01 Stability + performance,02 A redesigned UI,03 AI Motion Graphics) tucked cleanly into the bright brick wall space to the right of the speaker so neither his face nor the lower-third captions collide.

Step 6. Quote and generate one AI B-Roll cutaway
For the AI Motion Graphics line at 0:42.5, request a cutaway and a quote before AI video generation:
You: Generate one short B-roll shot that fits what I'm describing and place it over one line. Tell me the credit cost first.
Claude: Estimated cost: 4.38 ChatCut credits for one 4-second, 720p, 16:9 Seedance 2.5 shot. The final cutaway slowly pushes toward a monitor on a wooden desk in a warm-lit attic: a large orange square on dark gray above five blocks in cyan, purple, green, orange, and blue. This is an abstract illustration, not a real software interface or a complete editing timeline. Placed it over 0:42.5–0:46.5 while keeping your voice and captions underneath. The final clip cost 4.38 credits; the original attempt and two retries to remove garbled screen text cost 13.14 credits in total.
For the original master run, the generation estimate was 4.38 credits for one 4-second, 720p, 16:9 Seedance 2.5 clip (within the 5-credit cap for a single attempt; if your quote exceeds your cap, ask for a fresh quote for a still image with a slow push-in, or use existing footage). Submitting, waiting for the cloud render, and placing the silent cutaway fullscreen from 42.5 to 46.5 seconds took 5 minutes 29 seconds in that original run; this timing excludes the later retries. The clip used in the finished edit cost 4.38 ChatCut credits. Removing garbled screen text took three generations in total—the original attempt plus two retries—costing 13.14 credits. The first retry still had tiny garbled menu and track labels and failed frame-by-frame review; the second used only abstract colored blocks with no screen text and passed. The final screen is neither a real software interface nor a complete editing timeline.
Step 7. Add a calm music bed with automatic voice ducking
Add a restrained instrumental bed with automatic ducking:
You: Add a calm music bed and duck it under my voice.
Claude: Placed a 60-second CC BY 4.0 ambient track ('Relaxing Ambient Melodies' by ViraMiller) with a 1.5s fade-in, 2.0s fade-out, and automatic voice ducking (-20.3 LUFS integrated mix, -2.0 dBFS peak).
In the master run, 50 seconds (0 ChatCut credits) covered adding a free Creative Commons track—Relaxing Ambient Melodies by ViraMiller (CC BY 4.0)—trimmed it to 60.0 seconds with a 1.5-second fade-in and 2.0-second fade-out, and bound the dialogue track as anchor and the music track as follower (leaving the 0:36.4 whoosh independent). Across the full 60 seconds, the rendered mix measured -20.3 LUFS integrated loudness and -2.0 dBFS peak (-24.7 dBFS voice RMS vs. -27.3 dBFS music RMS).
Step 8. Export the 1080p 16:9 MP4 and build a 9:16 vertical Short from the same edit
Finally, export the horizontal master locally, then duplicate the approved timeline into a 9:16 vertical Short without touching the 16:9 edit:
You: Export a 1080p 16:9 MP4.
Claude: Exported final-16x9.mp4 locally (60.0s, 1920×1080, 30 fps H.264/AAC, 94.6 MB).
You: Now make a 9:16 vertical version for Shorts from the same edit, keeping the speaker centered, and export it too.
Claude: Duplicated the edit to a 'Shorts · 9x16' timeline, centered the speaker, resized the opening title and three-point list to fit 9:16, framed the wide monitor B-roll over a darkened background layer so the screen edges stay intact, and exported final-9x16.mp4 (1080×1920, 60.0s).
In the master run, both local renders and the vertical re-layout finished in 1 minute 50 seconds at 0 ChatCut credits (Desktop local export renders directly on your Mac or Windows GPU/CPU without using Free cloud-export minutes). Instead of blindly center-cropping every layer—which would clip a wide 16:9 title or chop off both sides of the B-roll monitor—the vertical pass repositioned the typography for 1080×1920 and placed the full 16:9 monitor shot over a darkened (-3 EV), scaled background copy:
Here is the 720×1280, 30 fps web preview (7.51 MB) of the vertical Short, encoded from the 1080×1920 local master (92.23 MB):
What did the shared master run take and cost?
| Step | Timeline action in the shared master | Recorded step duration | ChatCut credits including B-roll retries |
|---|---|---|---|
| 1. Import & transcribe | Imported 168.2s 1080p clip + ran word-timed cloud speech transcription | 0:38 | 0.00 |
| 2. 60s rough cut | Cut 7 fillers, retakes, and 3 long pauses → exact 60.0s (8 segments) | 1:33 | 0.00 |
| 3. Word-level captions | Arial 800 lower-third text + active-word #FF9B40 orange template + brand spelling checked | 1:06 | 0.00 |
| 4. Zooms & whoosh | Two 1.18× spring punch-in zooms + library Airy Short Whoosh (-14 dB) | 0:25 | 0.00 |
| 5. Motion Graphics | 0–3s "Make ideas move." title + 7.5–13.5s "THREE IMPROVEMENTS" list | 1:51 | 0.00 |
| 6. AI B-Roll cutaway | Quoted & generated 4s 720p Seedance 2.5 abstract monitor shot (42.5–46.5s); two later retries removed garbled screen text | 5:29 (original run only) | 4.38 final clip; 13.14 across 3 generations |
| 7. Music & ducking | 60s CC BY 4.0 ambient track, 1.5s/2.0s fades, voice-anchored auto-ducking (-20.3 LUFS) | 0:50 | 0.00 |
| 8. 16:9 + 9:16 exports | Exported 1920×1080 MP4, built reframed 9:16 timeline, exported 1080×1920 MP4 | 1:50 | 0.00 |
| Sum of step durations | Eight recorded step intervals | 13:42 | 13.14 including retries |
| Elapsed master-run window | 12:01:30–12:15:16, including 4 seconds between steps | 13:46 | 13.14 including retries |
These are the original shared master project’s recorded timings, including tool waits and checks; preparation, post-export verification, and later B-roll retries are outside this window. The credit column includes all three B-roll generations (13.14 credits total), while the clip used in the finished edit cost 4.38 credits; all other steps remained at 0 credits. They are not a Claude speed benchmark. The master agent used ChatCut’s tools for import, transcription, timeline edits, captions, graphics, stock-media search, generation quotes, previews, billing checks, and export. For continued editing in Adobe Premiere Pro or DaVinci Resolve, see the FCP7 XMEML handoff guide.
How can Claude Code use your local folder, CLAUDE.md, and skills?
Sending individual prompts works in any connected agent—for comparison, see our guide on how to edit video with ChatGPT. A useful Claude Code video editing workflow is to give a Local Code session access to the files beside your footage:
- Cut camera takes against a real
script.mdor show-notes file. When you start a session in your video folder (or click+→ Add folder in the Claude desktop app’s Code tab to attach a docs folder), Claude can compare what you planned to say inscript.mdagainst what you actually recorded, reorder out-of-sequence takes, and flag any missing line without inventing words:
- Save house packaging rules in
CLAUDE.mdor~/.claude/skills/. Instead of repeating your caption font, brand hex code, loudness target, and credit cap in every video session, drop a shortCLAUDE.mdinto your production folder—or save a reusableSKILL.mdprocedure in~/.claude/skills/(Claude Code skills guide) that applies across both the desktop Code tab andclaudein your terminal:
Which prompts can you copy for common Claude video editing tasks?
Paste any of these briefs into a connected Claude Code session on macOS or Windows. For dozens more templates by stage, see our video editing prompts library.
Clean up a talking-head video for YouTube (0 ChatCut credits)
Turn a long podcast or interview into 9:16 Shorts
For multi-clip vertical repurposing, combine this prompt with our walkthrough on turning long videos into Shorts:
Cut a screen-recorded product demo with callouts
Revise one cut or check an in-flight generation safely
When a cloud B-roll render takes a couple of minutes—or when one cut feels slightly rushed—point Claude at the exact timestamp or existing job rather than resending the whole brief:
What do you still need to check yourself before publishing?
The master run includes sampled visual checks and measured audio levels, but no completed end-to-end subjective listening review or final user acceptance. Before publishing, make these five checks:
- Brand names and capitalization in captions. Check names, acronyms, and casing in the caption layer before export. Keep caption corrections separate from edits to the spoken cut.
- Generated screens and signs. AI generation models often produce garbled lettering: specify
no text, no letters, no logos, inspect the result frame by frame, and budget credits for retries. Our final monitor cutaway passed that check and shows only an orange square and five colored blocks on dark gray, not a real software interface or a complete editing timeline. Use real screen recordings when viewers need to read an interface. - 9:16 vertical overlay bounds. A 9:16 crop can cut off wide headlines and monitor edges. Always scrub entry, hold, and exit frames on your vertical timeline to confirm titles, numbered lists, and wide inserts were resized cleanly.
- Headphone check on breaths and ducking. The master mix measured -20.3 LUFS integrated and -2.0 dBFS peak. Those readings do not by themselves prove ducking behavior, meet every delivery target, or tell you whether a tightened pause clipped an inhale or whether a word splice feels abrupt. Play the exported
.mp4with headphones from 0:00 to 1:00. - Music attribution and synthetic-media disclosure. Our edit used ViraMiller’s Relaxing Ambient Melodies under CC BY 4.0, which requires appropriate credit, a license link, and an indication of changes. For this use, credit the creator and source and note the 60-second excerpt, fades, and ducking. YouTube requires disclosing realistic synthetic scenes under YouTube’s altered or synthetic content policy.
How do you troubleshoot Claude Code and ChatCut connection issues?
Claude says it has no ChatCut tools after installation
Follow ChatCut’s connection instructions: finish sign-in, then open a new Code session; if needed, quit and reopen the agent app. This is the supported setup path for this connection, not a claim that all Claude plugins can only load at startup. You can inspect MCP registration with claude mcp list, or open Agents on this computer in the ChatCut Desktop sidebar and check the Claude Code connection row.
You pasted the prompt into claude.ai, mobile Claude, or the Chat / Cowork tab
ChatCut Desktop connects locally on your computer; it cannot install inside claude.ai, the iOS/Android apps, or the desktop app’s Chat and Cowork tabs. Switch to the Code tab in the Claude desktop app or run claude in Terminal or PowerShell.
The Code tab is open, but Claude still cannot reach your local project
Check the Environment selector directly below the prompt box in the Code tab (Claude Desktop docs). If it is set to Cloud or a remote SSH host, Claude is running on another machine and cannot talk to ChatCut Desktop on yours. Switch the environment to Local and start a new session.
Claude Code bills API credits instead of your Pro or Max subscription
Per Anthropic’s subscription setup guide, an active ANTHROPIC_API_KEY environment variable overrides your monthly Claude plan and routes turns to pay-as-you-go API billing. Remove ANTHROPIC_API_KEY from your shell profile (.zshrc, .bashrc, or system environment variables), restart Claude Code, and sign in with your claude.ai account.
You are running Linux
ChatCut Desktop supports macOS 13+ and Windows 10+ (x64), not Linux. Use a supported computer for this local workflow. The browser-based ChatCut web editor is a separate route; this guide does not verify a Linux hosted-plugin setup.
How was this guide produced?
We publish this first-party tutorial at ChatCut. All Claude Code setup details, Code tab environment controls, model aliases (opus / Opus 5.5, sonnet / Sonnet 5.5, haiku / Haiku 5.5, opusplan), ~/.claude/skills/ paths, and subscription rules were verified against Anthropic’s official Claude Code and Help Center documentation and ChatCut’s installation guides (chatcut.io/desktop, /claude-code-plugin, and /docs/product-facts) on November 18, 2026.
The eight-step attic-talk edit, elapsed master-run window (13 minutes 46 seconds, including gaps between Steps 1–8 but excluding later B-roll retries), credit settlement (4.38 ChatCut credits for the final 4-second 720p Seedance 2.5 cutaway; 13.14 credits across the original generation and two retries to remove garbled screen text; 0 credits for all other steps), preview stills, and both 60.0-second 1080p master MP4s come from our shared November 18, 2026 ChatCut Desktop master project, verified by full ffprobe/ffmpeg decode and sampled frames. Claude Code can access the same ChatCut editing tools, but the master run does not establish identical decisions, speed, or output across agents; the conversation cards above recreate the Claude desktop Code tab UI for visual clarity rather than claiming to be live window captures.
Frequently asked questions about editing video with Claude
How do I connect Claude to ChatCut to edit my own videos?
Does this work on claude.ai, on my phone, or in the Chat and Cowork tabs?
What is the difference between Claude Motion and Claude Code with ChatCut?
Which Claude plan and model do I need for video editing?
Does editing video with Claude Code spend ChatCut credits?
Can I adjust cuts, captions, and Motion Graphics manually after Claude finishes?
Does ChatCut Desktop work with Claude Code on Linux?
Using a different agent? The same ChatCut edit works from ChatGPT, Codex, Cursor, Antigravity, Grok Bot, OpenClaw, Hermes Agent, Meta Muse and WorkBuddy; each guide covers that agent’s install and quirks.