Skip to main content
Glama

Talk-n-Edit

The AI video editor you talk to. Tell Claude what you want, in plain words, and it cuts, captions and polishes your Reels, Shorts and TikToks on your own Mac.

Tests macOS License: MIT

You:     Make 3 Reels from the podcast in my inbox. Punchy style, Hinglish captions,
         zoom in on the big lines, and end each one with a follow button.

Claude:  Quick questions first: 9:16 at around 30 seconds each? Word-pop captions?
         Any music?  …  Here are 3 drafts. Reel 1 opens on "Nobody tells you this…"
         with a snap zoom; Reel 2 …

You:     Make the captions bigger and drop the emoji in Reel 2.

Claude:  Done. Finals are in your Clips folder.

No timeline to learn and no templates to fight. You direct, Claude edits, and you revise by just saying what to change.


Why Talk-n-Edit

  • You direct, it edits. Describe the Reel you want, look at the draft, ask for changes. Claude remembers the edit and changes only what you asked.

  • Anything you can describe. Transitions, zooms on a specific word, freeze frames, colour looks, blur over a logo, picture-in-picture, sound effects, animated titles and emoji. If something isn't built in, Claude writes a custom animated graphic and shows you a preview first.

  • Your language. Word-by-word captions in English, Hindi, Punjabi and more, including Hinglish (Hindi in Roman script) and translations.

  • Private. Your videos never leave your Mac. Transcription, face tracking and rendering all run locally.

  • No per-video fees. It uses the Claude app you already have. There are no credits or watermarks.

  • Safe with your files. Originals are never changed, and Talk-n-Edit never deletes anything.

Related MCP server: fcp-mcp-server

Install

You need: a Mac (Apple Silicon recommended), the Claude desktop app, about 6 GB of free space and an internet connection for the first setup.

1. Paste this into Terminal (press ⌘ Space, type Terminal, press Enter):

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/deepak199922/talk-n-edit/main/install.sh)"

It downloads Talk-n-Edit to ~/talk-n-edit and sets everything up: ffmpeg, Node, the speech model and the Claude connector. If something is missing, it tells you exactly what to do. The first run takes 5–15 minutes, mostly downloads.

2. Restart Claude. Quit it completely (⌘Q), then open it again.

3. Add the skill. The skill teaches Claude how to edit well. In Claude, open Customize → Skills → "+" → Create skill → Upload a skill and pick talk-n-edit-skill.zip from your Downloads folder. The installer created it and opens Finder on it.

4. Try it. Drop a video into Downloads/Talk-n-Edit/Inbox and ask Claude:

Make 3 Reels from the video in my Talk-n-Edit inbox.

Things to ask for

  • "Cut this podcast into 5 Shorts, each with a strong hook. Remove the pauses."

  • "Hype style: fast cuts, whip transitions with whooshes, word-pop captions, a 🔥 when I say the price."

  • "Cinematic look, slow motion on the intro, blur the number plate at 0:12."

  • "Show the product photo in a corner while I talk about it, then a follow button at the end."

  • "Captions in Hinglish, bigger, yellow on the key words."

  • "Make a custom animated price tag that spins in." (You'll see a preview before it's used.)

  • "Save this as my style." and next time: "Use my style."

What it can do

Cutting

Picks clips by sentence, removes pauses, speed ramps, freeze frames, reverse, stabilise, beat-synced cuts

Framing

16:9 → 9:16 / 1:1 / 4:5 that follows faces, split screen for two people, widescreen-film layout with the letterbox bars trimmed

Captions

Word-by-word karaoke, word-pop, clean or boxed styles; auto-emphasis; Hindi, Punjabi, Hinglish, translations; adjustable size

Effects

22 transitions, keyframed zooms and punch-ins, 14 colour looks plus your own LUTs, blur/pixelate masks, picture-in-picture and B-roll

Graphics

Animated titles, emoji, arrows, callouts, number count-ups (₹2,00,000 style), progress bar, follow/subscribe button, confetti, checklists, plus custom graphics on request

Audio

Your music, ducked under speech automatically; 11 built-in sound effects; voice clean-up; loudness set for Instagram and YouTube (−14 LUFS)

Quality checks

Every final is checked for black frames, dead air, loudness, hook timing and caption placement before you get it

Styles

Clean creator, hype, cinematic, podcast, minimal, product showcase, and your own saved styles

How it works

flowchart LR
  you["You: what you want, in plain words"] --> claude["Claude + Talk-n-Edit skill"]
  claude -- "plans the edit as a timeline" --> server["Talk-n-Edit on your Mac"]
  server --> analyse["Transcript (Whisper), faces, shots, beats"]
  server --> render["Render (FFmpeg + Remotion) and quality check"]
  render --> clips["Clips folder"]
  clips -- "draft or final" --> you

Claude decides what the video should be: which moments, which hook, which effects. Talk-n-Edit does the heavy lifting on your Mac: transcription, face tracking, rendering and checks. Each edit is saved as a version, so "go back to the previous one" just works.

Where your files go

Folder

What's in it

~/Downloads/Talk-n-Edit/Inbox

Drop your videos here

~/Downloads/Talk-n-Edit/Clips

Finished videos land here

~/TalkNEdit

Working files: copies of your imports, drafts, renders, saved styles, custom graphics

~/talk-n-edit

The app itself

Talk-n-Edit never deletes anything. Old drafts stay in ~/TalkNEdit/projects until you remove them.

Updating

Run the same install command again. It pulls the latest version and re-runs the setup. Then restart Claude.

Troubleshooting

Problem

Fix

Claude doesn't seem to have the tools

Quit Claude completely (⌘Q, not just closing the window) and reopen it. Still missing? Run the install command again.

You want a health check

Ask Claude: "Run the Talk-n-Edit system doctor." It lists anything that's wrong and how to fix it.

"ffmpeg is missing filters: subtitles"

Run brew install ffmpeg-full, then the install command again, then restart Claude.

A window asks to install "command line developer tools"

Click Install, wait for it to finish, then run the install command again.

The transcript is in the wrong language

Tell Claude the language: "It's in Hindi." It re-transcribes with that setting.

The first transcription is slow

The speech model downloads once (about 1.6 GB). Later runs are fast.

Animated graphics fail

Run the install command again. It downloads the small headless browser the graphics need.

Still stuck? Open an issue and paste the system doctor output.

Uninstalling

Talk-n-Edit never deletes files, so removing it is up to you:

  1. In Claude, remove the skill (Customize → Skills).

  2. Remove the "talk-n-edit" entry from ~/Library/Application Support/Claude/claude_desktop_config.json (setup left a backup copy next to it), then restart Claude.

  3. Delete the folders ~/talk-n-edit and ~/TalkNEdit, and ~/Downloads/Talk-n-Edit once you've kept the clips you want.

  4. Optional: brew uninstall ffmpeg-full uv, and delete the speech model in ~/.cache/huggingface/hub.

FAQ

How is this different from OpusClip and similar tools? Those are clip factories: upload, and the AI picks clips from templates. Talk-n-Edit is an editor you direct. You get exactly the edit you describe, revise it in plain words, and keep your videos on your own machine. If you need fully automatic bulk clipping with scheduling, a clip factory may suit you better.

What does it cost? Talk-n-Edit is free and open source. It runs through your Claude app, so it uses your normal Claude plan.

Which Claude plan do I need? Any plan that runs the Claude desktop app with local connectors. Uploading the skill needs a plan with Skills; see Anthropic's help article. Without the skill, Talk-n-Edit still works, but Claude edits less skilfully.

Can I post videos made from other people's content? Only if you have the rights. Talk-n-Edit edits whatever you give it, but clips from films, TV or other creators can get claimed or taken down.

Windows? Linux? Not packaged yet. Linux works for developers (see below). Windows is on the roadmap.

Can I use it with Claude Code or another MCP client? Yes. It's a standard MCP server. For example: claude mcp add talk-n-edit -- node ~/talk-n-edit/dist/index.js

Other systems (for developers)

On Linux, install ffmpeg (with libass), Node ≥ 22.13 and uv, then:

git clone https://github.com/deepak199922/talk-n-edit.git && cd talk-n-edit
npm install && npm run build
cd python && uv venv --python 3.12 .venv && uv pip install --python .venv/bin/python -r pyproject.toml && cd ..
node dist/index.js --check      # health check
npm test                        # end-to-end tests (~10 min)

Then add node /path/to/talk-n-edit/dist/index.js to your MCP client and upload skills/talk-n-edit (zipped) as a skill.

More detail is in the developer reference: every tool, the timeline format, the command line, brand kits and settings.

Contributing

Ideas, bug reports and pull requests are welcome. Start with CONTRIBUTING.md.

License and credits

Talk-n-Edit is MIT licensed. It builds on FFmpeg, Whisper, OpenCV, PySceneDetect, Remotion and the Model Context Protocol. See THIRD_PARTY_NOTICES.md. Note: Remotion (used for animated graphics) is free for individuals and companies of up to 3 people; larger companies need a Remotion company license.

Available Tools

18 tools
analysis_getA
Read-only

Read analysis results: scenes (each shot with start/end, face coverage, face position and size — use it to pick shots), beats (tempo + beat times for beat-synced cuts), faces (same per-scene view).

Inputs: kind (required), assetId (required).

analyzeA

Analyse an asset. transcript = word-level speech-to-text (clip picking, captions, silence removal); faces = face positions (auto-face / split reframing); scenes = shot boundaries; beats = music tempo + beat times (beat-synced cuts). Cached; returns a jobId.

Inputs: assetId (required), include (required), wait, force, +3 more.

catalog_listB
Read-only

List what timelines can use: components (text overlays + animated graphics + custom fx, with prop schemas), styles (saved edit styles), transitions, looks, sfx, music, luts, captionPresets, brandKits, fonts.

Inputs: kind (required).

frames_getA
Read-only

Grab up to 12 frames (as images) from a video asset or a finished render at the given times in seconds. Use it to check framing, hooks and cuts.

Inputs: times (required), width, assetId, renderId.

fx_createA

Create or update a custom animated effect when the built-in components can't do what the user wants. Write a React component (TSX) for Remotion:

  • Only import from "react" and "remotion" (useCurrentFrame, useVideoConfig, interpolate, spring, Easing, AbsoluteFill, Sequence…). No network, no require.

  • export default function Fx(props) {...}; draw on a TRANSPARENT background; size everything relative to useVideoConfig().width/height.

  • props.brand has colors {text, highlight, emphasis, outline, box}; font families 'VS Heading', 'VS Body', 'VS Deva', 'VS Guru' are loaded. Returns a 4-frame preview image (or the build/render error to fix). Show the preview to the user before using it as overlay component "fx:".

Inputs: code (required), name (required), description (required), preview, +4 more.

job_cancelC

Cancel a queued or running job.

Inputs: jobId (required).

job_statusA
Read-only

Status of one or more jobs (import, analyze, render, qa). Waits up to wait seconds for them to finish. A finished qa job includes its contact sheet and cut frames as images.

Inputs: jobIds (required), wait.

media_importA

Import video, audio or image files (or every media file in a folder) into a project. Copies originals into the workspace and builds a 540p proxy. Returns a jobId.

Inputs: paths (required), projectId (required), wait.

media_inspectB
Read-only

Probe an asset (duration, size, fps, audio, analyses done) and return a 4x3 contact sheet image of evenly spaced frames.

Inputs: assetId (required), contactSheet.

open_in_playerA

Open a finished render in the default video player (QuickTime on a Mac) so the user can watch it.

Inputs: renderId (required).

project_openB

Open a project by name (creates it if missing) and list its assets, analyses, timelines and renders. With no name, lists projects.

Inputs: name, create.

qa_checkA
Read-only

Check a finished render: format, duration, black frames, dead air, loudness, hook timing; returns a contact sheet and frames after the first cuts. Returns a jobId.

Inputs: renderId (required), wait.

render_startA

Render a timeline to MP4. quality 'draft' = fast 540-wide preview; 'final' = full quality, loudness-normalised, also copied to the export folder. Give timelineId (+version) or an inline timeline. Returns renderId + jobId.

Inputs: wait, quality, version, timeline, +1 more.

style_saveB

Save the user's editing preferences as a named style (look, transition, autoPunch, captions, sfxOnTransitions, hook, removeSilences, music volume, notes). Use it in timelines as "style": "" so every future edit matches.

Inputs: style (required).

system_doctorA
Read-only

Check ffmpeg, the Python analysis environment, fonts and folders. Run this first if anything fails.

Inputs: no named parameters.

timeline_getA
Read-only

Read a saved timeline (latest or a given version) so you can revise it: change what the user asked, then timeline_validate saves the next version.

Inputs: name, version, projectId, timelineId.

timeline_validateB

Validate a timeline JSON and save it as a new version. Returns resolved duration per clip, the spoken text each clip will contain, warnings, and errors with JSON paths. Same name + project continues the same timeline id.

Inputs: timeline (required), save.

transcript_getA
Read-only

Read an asset's transcript as numbered sentences with times: [s0012 03:41.2–03:47.9] text. Use sentence ids in clips ({sentences:{from,to}}) or the times as in/out. format 'words' gives per-word times.

Inputs: assetId (required), end, page, start, +1 more.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    B3.4/5.0

    Scored across 18 tools

    Disambiguation4/5

    Most tools target a distinct resource+action and are well separated (import vs inspect vs analyze vs render vs qa). The main near-overlaps are analyze vs analysis_get and frames_get vs media_inspect/analysis_get frame output, though the descriptions clarify trigger-vs-read and the differing frame purposes.

    Naming Consistency4/5

    The dominant pattern is resource_verb (media_import, timeline_validate, render_start, job_status, style_save, fx_create), used consistently across most of the surface. Deviations like the bare verb 'analyze', 'system_doctor', and 'open_in_player' are minor and still readable.

    Tool Count4/5

    18 tools is slightly heavy but each maps to a real stage of the video-editing pipeline (setup, ingest, analysis, authoring, render, QA, delivery). No tool appears clearly redundant, though the ceiling is near the upper bound for comfortable scoping.

    Completeness4/5

    The lifecycle is well covered end to end: doctor, project open, import, analyze, transcript/frames read, catalog, timeline validate/get, render, job status/cancel, QA, style save, fx create, and player open. Minor gaps exist around deletion/cleanup of projects and assets and listing renders, which agents can mostly work around.

    Maintenance

    ActivityMaintained
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • F
      license
      Not graded
      quality
      A
      maintenance
      Enables Claude to read, edit, and generate real Final Cut Pro timelines from exported FCPXML — trimming, splitting, silence removal, markers, roles, keyframes, transcript-based editing, and cross-NLE export — with journaled, undoable writes. It can also push an FCPXML straight into a running Final Cut Pro with zero clicks and render the timeline inline as structured data.
      -
    • F
      license
      Not graded
      quality
      A
      maintenance
      Enables Claude to read, edit, and generate real Final Cut Pro timelines from exported FCPXML, covering cut detection, trims, splits, silence removal, markers, roles, transcript-based edits, keyframes and speed ramps, plus live pushes into a running Final Cut Pro. Every write is journaled, undoable, and guarded by SHA-256 checks so edits can be reviewed and delivered safely.
      -