diff --git a/.releaserc.json b/.releaserc.json
index 553148a..778a55b 100644
--- a/.releaserc.json
+++ b/.releaserc.json
@@ -7,7 +7,8 @@
"preset": "angular",
"releaseRules": [
{ "type": "chore", "scope": "deps", "release": "patch" },
- { "type": "chore", "scope": "security", "release": "patch" }
+ { "type": "chore", "scope": "security", "release": "patch" },
+ { "type": "feat", "release": "patch" }
]
}
],
diff --git a/README.md b/README.md
index 05f0790..92f15a0 100644
--- a/README.md
+++ b/README.md
@@ -12,7 +12,7 @@ The open-source AI video editor for Claude Code and any MCP client: remove silen
-
+
@@ -40,12 +40,13 @@ Version 0.5.0 turns a raw talking-head recording into a clean, captioned cut: ve
| **Captions** | Burned-in captions timed to the words, as short phrases or one word at a time | `clip timeline captions` |
| **Reframe** | Turns a landscape cut into 9:16, 1:1 or 4:5, filling the frame and choosing the part that shows; zoom and position per clip | `clip timeline reframe` |
| **Loudness** | Evens out the exported mp4's volume to a target (-14 LUFS for YouTube, Shorts, Reels and TikTok; -16 for podcasts), measured in two passes | `clip timeline loudness` |
+| **Shorts from a long recording** | Turns ranges of words from a long recording into a vertical short on a timeline of its own: captioned, at -14 LUFS, with the cuts already made on the recording; the long cut stays as it is | `clip timeline short` |
| **Cut check** | Flags cuts where a word may be clipped, with a contact sheet of the frames on each side | `clip timeline check` |
| **Export to Final Cut Pro and DaVinci Resolve** | FCPXML 1.10 project | `clip timeline export --format fcpxml` |
| **Export to Premiere Pro** | Final Cut Pro 7 XML project, captions as `.srt` files | `clip timeline export --format xmeml` |
| **Render** | mp4 of the finished cut | `clip timeline export` |
-All of it works from chat (the Claude Code skill or the MCP server), the CLI, and the browser UI; the cut check isn't in the UI yet. Every edit is one step on a non-destructive timeline that `undo` and `redo` walk through. Underneath sit [19 single-file FFmpeg tools](#all-19-tools) for everything else: trim, split, concat, transitions, text and title cards, zoom/pan, overlay, speed, color, chroma key, stabilize, and render.
+All of it works from chat (the Claude Code skill or the MCP server), the CLI, and the browser UI; the cut check and shorts aren't in the UI yet. Every edit is one step on a non-destructive timeline that `undo` and `redo` walk through. Underneath sit [19 single-file FFmpeg tools](#all-19-tools) for everything else: trim, split, concat, transitions, text and title cards, zoom/pan, overlay, speed, color, chroma key, stabilize, and render.
---
@@ -157,6 +158,33 @@ A cut splits the clip into pieces: the first keeps its id, and `clip timeline sh
---
+## Quick start: a long recording → vertical shorts
+
+Each short is a timeline of its own, cut from the main timeline's clips, so the pauses and retakes you cut there stay cut. In Claude Code or Claude Desktop, ask:
+
+> *Transcribe podcast.mp4 and cut the pauses. Then find the three strongest moments, 20 to 60 seconds each, and make each one a short.*
+
+The same steps from the CLI:
+
+```bash
+clip timeline new
+clip timeline add-media podcast.mp4 # prints the clip id, e.g. clip_ec2f047d
+clip timeline transcribe clip_ec2f047d # transcribe it (once per file)
+clip timeline silence clip_ec2f047d # cut the pauses on main
+clip timeline transcribe --sentences | head -80 # one line per sentence: time, word indices, text
+clip timeline short best-1 clip_ec2f047d 1400 1478 --x 0.4 # words 1400–1478 as a 9:16 short
+clip timeline --timeline best-1 frame 3 check.jpg # look at it; move --x if the speaker is off
+clip timeline --timeline best-1 export best-1.mp4
+clip timeline list # main and every short
+clip timeline export --all ./shorts # every short, as ./shorts/.mp4
+```
+
+`short` cuts each edge inside the pause the audio shows there, keeping 0.1 s of room tone (where there's no clear pause it follows the transcript and warns), fills the 9:16 frame (`--x` picks the part that shows), sets -14 LUFS and burns in captions. Give several ` ` ranges to join moments, in the order you give them. Any command takes `--timeline ` to work on a short, `clip timeline --timeline best-1 undo` takes the short apart again, and `short best-1 … --replace` rebuilds it.
+
+In `clip ui`, select words in a clip's transcript and press **Make short**: the short opens in a tab of its own above the timeline, next to main. **Export all shorts** renders every one.
+
+---
+
## Another example: a product demo from screen recordings
Five screen recordings → one polished product demo. Type this into Claude Code (or Claude Desktop, via the MCP server):
@@ -324,7 +352,7 @@ No — nowhere. The CLI, the browser UI, the Claude Code skill, and the MCP serv
### Capabilities & limits
**Is it production-ready?**
-This release is feature-complete for local editing — a dark timeline editor (view, edit, preview, export), silence removal, transcript editing, captions, 9:16 / 1:1 / 4:5 reframing, loudness normalization, export to Final Cut Pro, DaVinci Resolve and Premiere Pro, 19 single-file tools, an MCP server, and 859 tests passing. The API surface is still pre-1.0 (tool schemas may change in minor ways before 1.0). Use it for real work; pin a version in CI.
+This release is feature-complete for local editing — a dark timeline editor (view, edit, preview, export), silence removal, transcript editing, captions, 9:16 / 1:1 / 4:5 reframing, loudness normalization, export to Final Cut Pro, DaVinci Resolve and Premiere Pro, 19 single-file tools, an MCP server, and 965 tests passing. The API surface is still pre-1.0 (tool schemas may change in minor ways before 1.0). Use it for real work; pin a version in CI.
**What's the maximum video size or duration?**
No hardcoded limit. FFmpeg streams the file rather than loading it into memory, and the browser UI streams uploads to disk — multi-GB recordings work fine. This release isn't tuned for projects with hundreds of clips or runtime above 30 minutes; for those, drive the CLI directly.
@@ -386,7 +414,7 @@ MakeMyClip Editor is built for that job: an open-source video editor that Claude
### Is there an MCP server for video editing?
-Yes. `npx -y @makemyclip/editor mcp` starts MakeMyClip Editor's local MCP server over stdio for Claude Desktop, Cursor, or any MCP client. Its tools show the timeline, edit it (add media, trim, split, remove silence, cut words, remove retakes, caption speech, reframe to 9:16, set loudness, transitions), check the cuts, transcribe, undo, redo, and export to mp4, FCPXML or Premiere XML. Media stays in one workspace folder, and no API key is needed.
+Yes. `npx -y @makemyclip/editor mcp` starts MakeMyClip Editor's local MCP server over stdio for Claude Desktop, Cursor, or any MCP client. Its tools show the timeline, edit it (add media, trim, split, remove silence, cut words, remove retakes, caption speech, reframe to 9:16, set loudness, make shorts, transitions), check the cuts, transcribe, undo, redo, and export to mp4, FCPXML or Premiere XML. Media stays in one workspace folder, and no API key is needed.
### How do I remove silence from a video automatically?
@@ -398,7 +426,11 @@ Yes. MakeMyClip Editor transcribes with whisper.cpp on your machine; install it
### Can Claude add captions or subtitles to a video?
-Yes. After transcription, MakeMyClip Editor's `clip timeline captions` (or the UI's Captions button) puts captions timed to the words on a track above the video, and export burns them in. Captions come as short phrases or one word at a time, with your choice of size, color, position, and background. Non-Latin scripts use an installed font that has the glyphs. In the Premiere export, captions become `.srt` files. On Linux, the bundled FFmpeg can't draw text, so set `MAKEMYCLIP_FFMPEG_PATH` to an FFmpeg built with libfreetype and libharfbuzz.
+Yes. After transcription, MakeMyClip Editor's `clip timeline captions` (or the UI's Captions button) puts captions timed to the words on a track above the video, and export burns them in. Captions come as short phrases or one word at a time, with your choice of size, color, position, and background. On a vertical canvas they sit above the bottom quarter, clear of the Shorts, Reels and TikTok buttons, and a line is never wider than fits at its font size. Non-Latin scripts use an installed font that has the glyphs. In the Premiere export, captions become `.srt` files. On Linux, the bundled FFmpeg can't draw text, so set `MAKEMYCLIP_FFMPEG_PATH` to an FFmpeg built with libfreetype and libharfbuzz.
+
+### Can it cut Shorts or Reels out of a podcast?
+
+Yes. Transcribe the recording, then `clip timeline short best-1 ` (or the `make_short` verb from chat) turns that range of words into a vertical short on a timeline of its own: 9:16, filling the frame, captioned, normalized to -14 LUFS, with the silence and retake cuts already made on the recording. Make as many as you like; the long cut stays untouched, and each short exports on its own. You or the agent pick the moments: nothing scores highlights for you, and nothing is uploaded.
### Can I finish an AI edit in DaVinci Resolve, Premiere Pro or Final Cut Pro?
@@ -422,7 +454,7 @@ auto-editor is a mature open-source command-line tool that cuts silence or motio
## Roadmap
-- **Now (v0.5.0)** — silence removal, local transcription, edit by transcript, retake removal, captions, 9:16 / 1:1 / 4:5 reframing, loudness normalization, click-free cuts, export to Final Cut Pro / DaVinci Resolve / Premiere Pro, on top of the dark timeline editor (view / edit / preview / export), 19 tools, MCP server, and snapshot/undo
+- **Now (v0.5.2)** — vertical shorts from a long recording, silence removal, local transcription, edit by transcript, retake removal, captions, 9:16 / 1:1 / 4:5 reframing, loudness normalization, click-free cuts, export to Final Cut Pro / DaVinci Resolve / Premiere Pro, on top of the dark timeline editor (view / edit / preview / export), 19 tools, MCP server, and snapshot/undo
- **Next** — model picker, SSE for live session updates, AI Elements / shadcn migration
- **1.0** — frozen tool schemas, MCP transport, published Anthropic skill, docs site
- **Beyond** — desktop app (Electron), cloud rendering via [MakeMyClip.com](https://makemyclip.com)
diff --git a/SKILL.md b/SKILL.md
index f28502d..8eadb11 100644
--- a/SKILL.md
+++ b/SKILL.md
@@ -1,6 +1,6 @@
---
name: editor
-description: Local, open-source video editor that agents drive with FFmpeg and whisper.cpp. Use when the user wants to edit a video file - cut out silence, pauses or dead air; remove repeated takes or cut words using the transcript; add captions or subtitles; make a vertical 9:16 (Shorts, Reels, TikTok), square or 4:5 version; even out the volume (loudness normalization); trim, split, join or crossfade clips; add text, title cards or zoom; clean up a talking-head video, screen recording, tutorial or podcast; render an mp4; or export the cut to Final Cut Pro, DaVinci Resolve or Premiere Pro (FCPXML / XML). Runs offline with no API key; every timeline edit can be undone; no generative AI.
+description: Local, open-source video editor that agents drive with FFmpeg and whisper.cpp. Use when the user wants to edit a video file - cut out silence, pauses or dead air; remove repeated takes or cut words using the transcript; add captions or subtitles; make a vertical 9:16 (Shorts, Reels, TikTok), square or 4:5 version; cut clips or shorts from a long video or podcast; even out the volume (loudness normalization); trim, split, join or crossfade clips; add text, title cards or zoom; clean up a talking-head video, screen recording, tutorial or podcast; render an mp4; or export the cut to Final Cut Pro, DaVinci Resolve or Premiere Pro (FCPXML / XML). Runs offline with no API key; every timeline edit can be undone; no generative AI.
license: MIT
compatibility: Requires Node 24+. FFmpeg is auto-downloaded via ffmpeg-static on first use; override with MAKEMYCLIP_FFMPEG_PATH. Transcription (transcript cuts, retakes, captions) needs a user-installed whisper.cpp and a model file set in MAKEMYCLIP_WHISPER_MODEL.
metadata:
@@ -22,6 +22,7 @@ Trigger this skill when the user asks to:
- Cut the silence, pauses, or dead air out of a recording ("tighten it up", "remove the gaps")
- Remove repeated takes or bad takes, or cut a sentence or words by the transcript
- Clean up a talking-head video, tutorial, lecture, or video podcast
+- Cut clips or vertical shorts (Shorts, Reels, TikTok) out of a long video, webinar or podcast
- Add captions or subtitles timed to the speech, or titles and lower-thirds
- Zoom, pan, or apply Ken Burns to a region of a clip
- Assemble multiple clips into one (concat, montage)
@@ -322,16 +323,18 @@ To cut the pauses out of a talking clip already on the timeline, use `timeline s
After cutting, run `timeline check` before exporting: it lists every hard cut whose audio is loud right at the join (usually a clipped word or a click) and writes a contact sheet (`--sheet out.jpg`) with the outgoing clip's last frame and the incoming clip's first for each cut. Open the sheet to see the jumps. A flagged cut is a reason to trim or move that boundary, not a verdict. Export fades the audio over 8 ms on each side of every hard cut, so a cut through steady sound doesn't click; the check still flags cuts in sound, since those can clip a word.
-To read what's said in a clip, run `timeline transcribe `: it returns the clip's words with their timeline times, so you can find a sentence before cutting around it. Transcription is optional and runs a local whisper.cpp; if it isn't set up, the command says how (install whisper.cpp, download a model, set `MAKEMYCLIP_WHISPER_MODEL`). The first run per source file takes a while; later runs reuse the stored transcript. Whisper tends to leave out filler words like "um", so don't rely on the transcript to find them.
+To read what's said in a clip, run `timeline transcribe `: it returns the clip's words with their timeline times, so you can find a sentence before cutting around it. For a long recording, read `timeline transcribe --sentences` instead: one line per sentence of the whole timeline (`12:34 #1400–1413 clip_ab12 text`, the time, the word indices, the clip it starts in), and `--from S --to S` for a stretch; pipe it through `head` to read a page at a time. Transcription is optional and runs a local whisper.cpp; if it isn't set up, the command says how (install whisper.cpp, download a model, set `MAKEMYCLIP_WHISPER_MODEL`). The first run per source file takes a while; later runs reuse the stored transcript. Whisper tends to leave out filler words like "um", so don't rely on the transcript to find them.
To cut by words, use the `index` values from `timeline transcribe`: `timeline cut-words ` removes those words (inclusive) and moves later clips on the track up. Add `--transcript ` from the same `timeline transcribe` output so the cut is refused, not misplaced, if the clip was transcribed again since. The transcript also lists `retakes`, phrases said twice in a row; `timeline remove-retakes ` cuts the earlier take of each (`--dry-run` lists them first). Each is one edit that `timeline undo` reverses. A cut splits the clip, so transcribe the resulting clips again before the next cut, and run `timeline check` afterwards.
-To caption speech, run `timeline captions` once the cuts are done: it turns the stored transcripts into text clips on a `captions` track above the video, timed to the words, and export burns them in. With no clip id it captions every transcribed clip and replaces the whole caption track; `timeline captions ` replaces only the captions over that clip. Captions are short phrases (up to 7 words, broken at pauses and sentence ends); `--words 1` shows one word at a time. Style with `--font --color --anchor --background ` (default white on a dark box, bottom center). Captions don't move with later cuts, so run it again after one. It's one edit that `timeline undo` reverses, and `timeline frame ` shows a captioned frame. Text outside Latin (Cyrillic, Chinese, Japanese, Korean) is drawn with an installed font that has its glyphs; if none does, export says which characters are missing — install a font (for CJK, Noto Sans CJK) or set `MAKEMYCLIP_FALLBACK_FONT` to a font file. Nothing is downloaded.
+To caption speech, run `timeline captions` once the cuts are done: it turns the stored transcripts into text clips on a `captions` track above the video, timed to the words, and export burns them in. With no clip id it captions every transcribed clip and replaces the whole caption track; `timeline captions ` replaces only the captions over that clip. Captions are short phrases (up to 7 words, broken at pauses and sentence ends); `--words 1` shows one word at a time. Style with `--font --color --anchor --background ` (default white on a dark box, bottom center). On a vertical (portrait) canvas captions sit above the bottom quarter and keep 14% in from each side, clear of the Shorts, Reels and TikTok overlays; `--margin ` sets the distance from the bottom (or top) edge. A line never gets wider than fits at its font size (22 characters at the default size on 1080x1920), so a bigger `--font` makes shorter lines. Captions don't move with later cuts, so run it again after one. It's one edit that `timeline undo` reverses, and `timeline frame ` shows a captioned frame. Text outside Latin (Cyrillic, Chinese, Japanese, Korean) is drawn with an installed font that has its glyphs; if none does, export says which characters are missing — install a font (for CJK, Noto Sans CJK) or set `MAKEMYCLIP_FALLBACK_FONT` to a font file. Nothing is downloaded.
To make a vertical (9:16), square (1:1) or feed (4:5) version, run `timeline reframe 9:16`. It changes the canvas shape, keeping its short side (1920x1080 becomes 1080x1920), and fills it: `--fit cover` (the default) crops what overflows, `--fit contain` shows the whole picture with bars. Look at `timeline frame ` first to see where the speaker is, then pass `--x` (0 shows the left edge of the source, 0.5 the center, 1 the right edge; `--y` likewise, top to bottom) to keep them in frame. `--x` sets every media clip on the timeline now; clips added later start centered. To frame one clip differently, run `timeline framing --x 0.3` (also `--scale` 1–4 to punch in, `--fit cover|contain|inherit`, `--reset`). Check with `timeline frame ` after; the `timeline check` contact sheet shows the source frames, not the crop. Captions keep their size, since the short side doesn't change. Each is one edit that `timeline undo` reverses. A reframed cut exports as mp4; the editor-project export refuses it.
To even out the volume of the exported mp4, run `timeline loudness -14` (YouTube, Shorts, Reels, TikTok) or `-16` (podcasts); `off` turns it off. Export then measures the finished mp4 and normalizes it to that integrated loudness in a second pass (EBU R128, true peak at most -1.5 dBTP), and reports `loudness` with the measured before and after values. `mode: dynamic` means loudnorm couldn't apply one gain (steady or very short audio, a wide loudness range, or peaks); `skipped` means the audio was silent or too short to measure. It's one setting on the timeline that `timeline undo` reverses, it applies to the mp4 only, and it doesn't even out clips recorded at different levels.
+To cut vertical shorts out of a long recording, transcribe it and make the cuts you want everywhere (pauses, retakes) on the main timeline first, then read `timeline transcribe --sentences`, pick each moment as whole sentences and run `timeline short ` with their word indices, where `` is a new name such as `best-1`. Pick whole sentences, 15–60 s in all; you choose the moments, the editor doesn't score them. The short is a timeline of its own, cut from main's clips (main's cuts carry in, main itself doesn't change): it cuts each edge inside the pause the audio shows there, keeping 0.1 s of room tone, so it never keeps the next sentence's first syllable, fills a 9:16 frame, sets -14 LUFS and burns in captions. More ` ` triples join several moments in the order given, so a later line can open the short as its hook. Options: `--x 0..1` to keep the speaker in frame, `--aspect 1:1|4:5`, `--words N` per caption, `--no-captions`, `--loudness `, `--source `, `--dry-run` (prints the pieces), `--replace` to rebuild it. Its `warnings` name a range that starts or ends mid-sentence (widen it to the sentence's edge unless the cut is meant) and an edge with no clear pause in the audio (music under the speech, or no break), which follows the transcript instead: look at that edge or listen to it. Every other command works on a short with `--timeline `: look at it with `timeline --timeline best-1 frame 3 check.jpg`, adjust with `timeline --timeline best-1 reframe 9:16 --x 0.3`, and render it with `timeline --timeline best-1 export best-1.mp4` (`timeline export --all ` renders every short as `/.mp4`). `timeline list` shows every timeline, `timeline --timeline best-1 undo` takes the short apart, and `timeline delete best-1` removes it. Over MCP, it's the `make_short` verb in `timeline_edit` with `timeline` set to the new id; every timeline tool takes `timeline`, and `timeline_show` lists them all.
+
To hand the cut to an editor app, export the timeline as an editor project instead of a render. `timeline export cut.fcpxml` (or `--format fcpxml`) writes FCPXML 1.10 for Final Cut Pro and DaVinci Resolve; `timeline export cut.xml` (or `--format xmeml`) writes Final Cut Pro 7 XML for Premiere Pro, which Resolve reads too. Premiere doesn't import FCPXML, so use `xmeml` for it. Both reference the source files by absolute path (nothing is copied, so keep the sources where they are), place camera files by their start timecode, and keep the mp4's timing: clips, gaps, fades, volume, fade/dissolve transitions (as cross dissolves) and black or colored cards. Text over the picture becomes titles in FCPXML; in the Premiere XML it goes to `.srt` caption files next to the XML, since Premiere drops that format's titles. What the file can't hold — speed, color adjustments, chroma key, a reframe's fill, zoom or position (set the timeline to `--fit contain` and reset clip framing to carry it), rotation, wipe/slide/circle transitions, audio-only tracks, media on two tracks, a frame rate editors don't offer — is refused with the fix to make; what it holds only in part (text styling, caption boxes) is listed in the output's `notes`.
### Open the local UI (implemented)
diff --git a/docs/backlog/caption-safe-zone.md b/docs/backlog/caption-safe-zone.md
deleted file mode 100644
index 7a7d784..0000000
--- a/docs/backlog/caption-safe-zone.md
+++ /dev/null
@@ -1,3 +0,0 @@
-# Caption safe zone for vertical video
-
-Captions sit 40 px from the bottom edge (`EDGE` in `src/timeline/compile.ts`) whatever the canvas shape. On a 9:16 export, the Shorts, Reels and TikTok overlays (caption text, buttons) cover roughly the bottom fifth of the frame, so bottom-anchored captions land under them. Raise the default caption position on portrait canvases, or add a safe-zone offset to the caption style.
diff --git a/docs/backlog/caption-width-follows-font-size.md b/docs/backlog/caption-width-follows-font-size.md
deleted file mode 100644
index 6b0e133..0000000
--- a/docs/backlog/caption-width-follows-font-size.md
+++ /dev/null
@@ -1,3 +0,0 @@
-# Caption width follows the font size
-
-`groupCaptions` breaks captions at 32 Latin characters (`MAX_CAPTION_WIDTH` in `src/timeline/captions.ts`) whatever the font size, and drawtext doesn't wrap. With the default size (the canvas short side / 20) a caption fits, but a larger `--font` makes long captions run off both edges. Derive the limit from the effective font size and the canvas width.
diff --git a/docs/backlog/caption-width-font-metrics.md b/docs/backlog/caption-width-font-metrics.md
new file mode 100644
index 0000000..b34fc58
--- /dev/null
+++ b/docs/backlog/caption-width-font-metrics.md
@@ -0,0 +1,3 @@
+# Caption width from the font's own metrics
+
+Caption lines are fitted by `textWidthEm` in `src/timeline/captions.ts`, an estimate per character class that errs wide for Arial and DejaVu Sans (W and m at 1 em, other capitals 0.8, non-Latin letters 0.95–1.1). A font wider than those, or a glyph outside the classes, can still overflow the safe area. Read the advance widths from the font file the caption is drawn with (`fontFileFor` in `src/ffmpeg/font.ts` already parses its character map) and fit lines by them.
diff --git a/docs/backlog/confine-render-media.md b/docs/backlog/confine-render-media.md
new file mode 100644
index 0000000..b5ab9c8
--- /dev/null
+++ b/docs/backlog/confine-render-media.md
@@ -0,0 +1,9 @@
+# Re-check media against the workspace on every render for the UI and MCP
+
+`POST /api/timelines/export` re-checks each short's files with `confinedMediaMap`, so a registered file swapped for a symlink out of the workspace is refused. Three older render paths still take the registry as ingested (`buildMediaMap`):
+
+- `POST /api/timeline/export` (mp4) in `src/ui/server.ts`;
+- `GET /api/timeline/frame` in `src/ui/server.ts`;
+- the mp4 path of MCP `timeline_export` in `src/mcp/server.ts`.
+
+The transcript and project-file paths already confine. Pass `confinedMediaMap(comp)` on these three, and test each with a symlinked file, as `tests/ui-timelines.test.ts` does for export-all. Found by the Codex review of the UI shorts PR (2026-11-15).
diff --git a/docs/backlog/cut-words-pause-edges.md b/docs/backlog/cut-words-pause-edges.md
new file mode 100644
index 0000000..5c88d73
--- /dev/null
+++ b/docs/backlog/cut-words-pause-edges.md
@@ -0,0 +1,3 @@
+# Cut words at the pause
+
+`cut_words` and `remove_retakes` cut from the first word's start to the last word's end as whisper.cpp gives them (`planWordCut` in `src/timeline/transcript.ts`). whisper gives each pause to the word before it and runs late at sentence edges. On a 30-sentence test recording, cutting a sentence would also remove up to 0.35 s of the next sentence's first syllable, at 18 of 29 boundaries. Place those cut edges in the pause the audio shows, as `make_short` does (`src/timeline/pauses.ts`), through `VerbContext.speechLevels`.
diff --git a/docs/backlog/frame-preview-short-window.md b/docs/backlog/frame-preview-short-window.md
deleted file mode 100644
index 177b62f..0000000
--- a/docs/backlog/frame-preview-short-window.md
+++ /dev/null
@@ -1,3 +0,0 @@
-# Frame preview encodes a short window
-
-`buildFrameAtPlan` (`src/timeline/compile.ts`) encodes the whole clip's segment to grab one frame, so every framing change in the browser UI re-encodes the full clip before the viewer updates. On a long uncut recording that takes seconds. Encode only a short window around the requested time (`-ss` before the input, `-t` of a few frames) and adjust the frame offset, keeping the same filter chain so the preview still matches export.
diff --git a/docs/backlog/retake-test-model.md b/docs/backlog/retake-test-model.md
new file mode 100644
index 0000000..16fe1aa
--- /dev/null
+++ b/docs/backlog/retake-test-model.md
@@ -0,0 +1,3 @@
+# Retake integration test depends on the model
+
+`tests/integration/timeline-transcript.test.ts` ("removes a repeated take as one undoable edit") runs whenever whisper.cpp and a model are set up. With `ggml-base.en.bin` it fails on `feat/shorts` too: the model's transcript of the test's spoken phrase has no immediate repeat, so `retakes` is empty. Make the test independent of the model (assert on a stored transcript with a known repeat, as the unit tests do) or pin the model it was written for.
diff --git a/docs/backlog/review-test-timeout.md b/docs/backlog/review-test-timeout.md
new file mode 100644
index 0000000..955a01f
--- /dev/null
+++ b/docs/backlog/review-test-timeout.md
@@ -0,0 +1,3 @@
+# Review integration test timeout
+
+`tests/integration/timeline-review.test.ts` ("flags a cut planted mid-tone…") runs under vitest's default 5 s timeout but renders several FFmpeg steps, so it sometimes times out on a busy machine (seen once in two runs on `feat/shorts` at 18511f1). Give it an explicit timeout like the other integration tests (`120_000`).
diff --git a/docs/backlog/segment-duration-rounding.md b/docs/backlog/segment-duration-rounding.md
new file mode 100644
index 0000000..46e8c92
--- /dev/null
+++ b/docs/backlog/segment-duration-rounding.md
@@ -0,0 +1,3 @@
+# Segment duration rounding
+
+Export renders each clip as its own segment and joins them, and each segment comes out a little longer than its clip: the video rounds up to a whole frame and the AAC audio to a whole 1024-sample frame. A three-piece short planned at 16.200 s exported as 488 video frames (16.267 s at 30 fps) and 16.25 s of audio. Check whether the audio-video offset accumulates over a timeline with hundreds of cuts (a silence-cut hour), and if it does, trim each segment to a frame-aligned length or carry the remainder into the next segment.
diff --git a/docs/backlog/timeline-color-validation.md b/docs/backlog/timeline-color-validation.md
deleted file mode 100644
index 643dac7..0000000
--- a/docs/backlog/timeline-color-validation.md
+++ /dev/null
@@ -1,3 +0,0 @@
-# Timeline color validation
-
-Timeline `color` / `background` fields are bare `z.string()` (src/timeline/composition.ts, ops.ts) and `quoteColor` in src/timeline/compile.ts passes non-`#` values into the filtergraph unescaped, so a value like `red,drawtext=...` adds a filter; validate them as hex or a named-color allowlist in the schema.
diff --git a/docs/backlog/ui-inspector-height.md b/docs/backlog/ui-inspector-height.md
new file mode 100644
index 0000000..4efeff5
--- /dev/null
+++ b/docs/backlog/ui-inspector-height.md
@@ -0,0 +1,5 @@
+# Clip inspector has no room on a short window
+
+In `clip ui`, the viewer takes up to 44vh plus its bar, and the clip inspector (`.stage-detail`) gets whatever is left of `.main`. At a 1000 px tall window, `.main` is 468 px and the viewer fills all of it, so the inspector is 0 px tall: a selected clip's transcript, framing and Make short can't be reached. It works at 1500 px. Seen on 2026-11-15 while checking the shorts tabs; the layout was the same before them, and the tabs bar takes another ~30 px once a short exists.
+
+Fix: give the inspector a minimum height (or let the viewer shrink first), and check at 900 px tall.
diff --git a/llms.txt b/llms.txt
index e245b0a..21f2713 100644
--- a/llms.txt
+++ b/llms.txt
@@ -9,18 +9,19 @@
- **Backend:** FFmpeg (downloaded by `ffmpeg-static` at npm install; the Claude Desktop extension uses an installed FFmpeg); optional user-installed whisper.cpp for transcription
- **Surfaces:** Claude Code skill · `clip` CLI · browser UI at `127.0.0.1:5573` · MCP server (`clip mcp`)
- **Agent access:** Claude Code skill + MCP server (`clip mcp`); no API key — the editor embeds no language model, it runs on whichever Claude you already use
-- **Tests:** 859 passing
+- **Tests:** 965 passing
- **License:** MIT (code); FFmpeg runs as a subprocess under its own license (GPLv3 builds from `ffmpeg-static`, except a nonfree Apple Silicon build; the `.mcpb` ships no FFmpeg)
## What v0.5.0 does
- **Remove silence:** `clip timeline silence ` cuts the pauses out of a recording (`--sensitivity`, `--dry-run`)
-- **Transcribe locally:** `clip timeline transcribe ` returns each word with its timeline time, via whisper.cpp on the user's machine
+- **Transcribe locally:** `clip timeline transcribe ` returns each word with its timeline time, via whisper.cpp on the user's machine; `clip timeline transcribe --sentences` lists a long recording as numbered sentences
- **Edit by transcript:** `clip timeline cut-words ` cuts a range of words
- **Remove repeated takes:** `clip timeline remove-retakes ` cuts the earlier take of each phrase said twice in a row
-- **Captions:** `clip timeline captions` burns in captions timed to the words (short phrases or one word at a time)
+- **Captions:** `clip timeline captions` burns in captions timed to the words (short phrases or one word at a time); on a vertical canvas they stay clear of the Shorts, Reels and TikTok overlays, and their width follows the font size
- **Reframe:** `clip timeline reframe 9:16` turns a landscape cut vertical (also 1:1, 4:5); `--x` picks the part that shows, `clip timeline framing ` sets one clip's zoom and position
- **Loudness:** `clip timeline loudness -14` normalizes the exported mp4 to -14 LUFS (or `-16` for podcasts) in two passes
+- **Shorts from a long recording:** `clip timeline short best-1 ` (MCP: the `make_short` verb) turns a range of transcript words into a 9:16 short on a timeline of its own, captioned and at -14 LUFS, keeping the cuts made on the main timeline; `--timeline ` points any command at a short, and `clip timeline export --all ` renders every short. In `clip ui`, Make short works from a transcript selection and each short gets a tab
- **Cut check:** `clip timeline check --sheet cuts.jpg` flags cuts that may clip a word and writes a contact sheet
- **Export to pro editors:** `clip timeline export --format fcpxml` (Final Cut Pro, DaVinci Resolve) or `--format xmeml` (Premiere Pro; captions as .srt)
- Every timeline edit is one undoable step (`clip timeline undo` / `redo`), from the CLI, the MCP server or the browser UI
@@ -52,7 +53,7 @@ Plus session meta-ops not in the tool registry: snapshot, undo, inspect, delete.
- Different category from Runway/Pika (generative AI), Premiere/DaVinci (heavy NLEs) — but it exports its cut to Final Cut Pro, Resolve and Premiere for finishing
- Overlaps with auto-editor (open-source CLI: silence cuts, transcript edits, NLE export); MakeMyClip Editor adds a shared timeline document that an agent and a person (browser UI) both edit, with undo per step
- Transcription is local (whisper.cpp); nothing is uploaded
-- Single workspace folder holds composition.json (the timeline), session.json (the op log), output files, snapshots/ — fully version-controllable
+- Single workspace folder holds composition.json (the main timeline), timelines/ (shorts and other named timelines), session.json (the op log), output files, snapshots/ — fully version-controllable
- Stream-copy for lossless trim/split/concat; re-encodes only when filter graphs require it
## Architecture
diff --git a/mcpb/manifest.json b/mcpb/manifest.json
index 2841441..dc6e9c5 100644
--- a/mcpb/manifest.json
+++ b/mcpb/manifest.json
@@ -40,7 +40,7 @@
},
{
"name": "timeline_edit",
- "description": "Apply editing verbs as one undoable change: add media, text or color, trim, move, split, remove, remove silence, cut words or retakes, caption speech, reframe to 9:16 / 1:1 / 4:5, frame a clip, set loudness, transition."
+ "description": "Apply editing verbs as one undoable change: add media, text or color, trim, move, split, remove, remove silence, cut words or retakes, caption speech, reframe to 9:16 / 1:1 / 4:5, frame a clip, set loudness, make a vertical short from transcript words, transition."
},
{
"name": "timeline_check",
diff --git a/src/cli-timeline.ts b/src/cli-timeline.ts
index e623c1e..41485d2 100644
--- a/src/cli-timeline.ts
+++ b/src/cli-timeline.ts
@@ -1,5 +1,11 @@
import { appendOp } from './session/store.js';
-import { CaptionError, DEFAULT_CAPTION_WORDS, planCaptions } from './timeline/captions.js';
+import {
+ CaptionError,
+ captionSideMargin,
+ DEFAULT_CAPTION_WORDS,
+ planCaptions,
+} from './timeline/captions.js';
+import { invalidColorMessage, isColor } from './timeline/color.js';
import {
buildFrameAtPlan,
checkExportable,
@@ -9,6 +15,7 @@ import {
import {
AnchorSchema,
type Clip,
+ type Composition,
clipDuration,
clipEndSec,
clipsAtTime,
@@ -24,11 +31,15 @@ import {
type TextStyle,
} from './timeline/composition.js';
import {
+ deleteTimeline,
+ listTimelines,
mutateComposition,
+ NoTimelineError,
readComposition,
readDocOpLog,
redoDocOp,
resetComposition,
+ TimelineIdError,
undoLastDocOp,
} from './timeline/document-store.js';
import {
@@ -39,6 +50,7 @@ import {
} from './timeline/interchange.js';
import { buildMediaMap, buildMediaRefs } from './timeline/media-registry.js';
import { colorClip, mediaClip, textClip } from './timeline/ops.js';
+import { measureSpeechLevels } from './timeline/pauses.js';
import {
AspectSchema,
type FramingFit,
@@ -48,13 +60,16 @@ import {
planReframe,
} from './timeline/reframe.js';
import { reviewCuts } from './timeline/review.js';
-import { runPlan } from './timeline/run-plan.js';
+import { exportTimelines, runPlan } from './timeline/run-plan.js';
+import { MAIN_TIMELINE, TimelineIdSchema } from './timeline/schema.js';
+import { checkShortTimelines, planShort, type ShortRange } from './timeline/shorts.js';
import { analyzeSilence, DEFAULT_SENSITIVITY, SensitivitySchema } from './timeline/silence.js';
import {
clipWords,
findRetakes,
planWordCut,
storedTranscriptForClip,
+ timelineOutline,
transcribeClip,
} from './timeline/transcript.js';
import { DEFAULT_VERB_TRACK as DEFAULT_TRACK, ensureTrack, trackEnd } from './timeline/verbs.js';
@@ -63,6 +78,9 @@ import { getWorkspace, newOutputPath, resolveInput } from './workspace.js';
const TIMELINE_HELP = `clip timeline — build and export a non-destructive composition
+Every command works on the main timeline unless --timeline names another
+(a short made with \`short\` is a timeline of its own).
+
clip timeline new [--width W --height H --fps F --background C --fit contain|cover]
clip timeline add-media [ ] [--start S] [--track T]
clip timeline add-text [--start S] [--track T] [--font N --color C --anchor A --background C]
@@ -74,13 +92,16 @@ const TIMELINE_HELP = `clip timeline — build and export a non-destructive comp
clip timeline remove
clip timeline silence [--sensitivity 0..1] [--dry-run]
clip timeline check [--sheet ]
- clip timeline transcribe [--model ] [--language ]
+ clip timeline transcribe [] [--sentences] [--from S] [--to S] [--model ] [--language ]
clip timeline cut-words [--transcript ]
clip timeline remove-retakes [--dry-run] [--transcript ]
- clip timeline captions [] [--words N] [--track T] [--font N --color C --anchor A --background C|none] [--dry-run]
+ clip timeline captions [] [--words N] [--track T] [--font N --color C --anchor A --background C|none --margin PX] [--dry-run]
clip timeline reframe <9:16|1:1|4:5|16:9|WxH> [--fit cover|contain] [--x 0..1] [--y 0..1] [--dry-run]
clip timeline framing [--fit cover|contain|inherit] [--scale 0.1..4] [--x 0..1] [--y 0..1] [--reset]
clip timeline loudness <-14|-16|LUFS|off>
+ clip timeline short [ …] [--source T] [--aspect 9:16|1:1|4:5|16:9] [--x 0..1] [--y 0..1] [--words N] [--no-captions] [--loudness LUFS|off] [--replace] [--dry-run]
+ clip timeline list
+ clip timeline delete
clip timeline show
clip timeline at
clip timeline frame [