You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Cut a clip with --start, --end and --duration.conv talk.mp4 --start 1:02 --end 1:10 writes talk-1m02s-1m10s.mp4, and an output name
changes the format as well (clip.gif, .mp3). Times are seconds, m:ss
or h:mm:ss, and a leading - counts back from the end. The cut is exact
to the frame. A cut that starts after 0 re-encodes the video, and so does
one from 0 of video with B-frames, with a note that says why; any other
cut from 0 is a stream copy. --max-size sizes the clip rather than the
whole file, and --strip-metadata, --fps, --resize and --crf work
as usual. A --to batch writes into a folder named after the range
(1m02s-1m10s/), and --json gets a range object.
conv trim FILE cuts clips in the terminal. The frame under the
slider is drawn at the top, with a bar of the video's colours and a bar of
its loudness below. The arrow keys move the slider, c marks a clip's
start and then its end, and w writes every clip the way the flags would.
A clip can keep the picture alone or the sound alone. The picture uses
kitty's or iTerm2's graphics where the terminal has them and text
characters elsewhere; --graphics overrides the choice.
Install convkit 0.5.0
Install prebuilt binaries via shell script
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/shdwfruit/convkit/releases/download/v0.5.0/convkit-installer.sh | sh
Location note. Phone photos and videos record where they were taken,
and a converted copy keeps that in most formats. conv now says so when it
applies: "The source records a GPS location, and the jpg keeps it; add
--strip-metadata to remove it." It reads an image's EXIF and a video or
audio file's tags. A WebP source, a location kept only in XMP, and a HEIC
photo on ImageMagick 6 get no note.
--strip-metadata removes the location and the rest of the metadata:
camera, serial number, owner, capture time. An image keeps only its colour
profile and is turned upright first. Video and audio keep their title,
artist, album, album artist, composer, genre and track and disc numbers,
and a stream copy stays a stream copy. conv photo.jpg --strip-metadata
writes photo-stripped.jpg; a jpg, webp or avif is re-encoded to do it,
and a note says so. Documents are not covered yet and refuse the flag.
Changed
Sized and stripped batches write into a folder. A --to batch with --max-size or --strip-metadata now writes into a folder next to its
files, named after the flag (clips/10mb/, photos/stripped/), unless -o says where. Both flags take files already in the target format as
inputs, so running the same batch again used to size or strip its own
results. Until now --max-size batches wrote beside their inputs. A
single file still gets clip-10mb.mp4 beside it.
Same-named inputs no longer stop a batch. Two inputs that would write
one file, like a.jpg and a.png with --to webp, keep their source
format in the name (a-jpg.webp, a-png.webp) instead of the whole batch
being refused.
Install convkit 0.4.0
Install prebuilt binaries via shell script
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/shdwfruit/convkit/releases/download/v0.4.0/convkit-installer.sh | sh
Notes only when they apply. Converting to jpg, png or bmp no longer
warns about transparency or extra frames the source doesn't have, so a
phone photo converts without a note. A GIF made from a video of 30 seconds
or less no longer gets the memory note. When conv can't tell, the note
still shows. warnings in --json follows the same rule, so it can now be
empty where it wasn't before.
Install convkit 0.3.1
Install prebuilt binaries via shell script
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/shdwfruit/convkit/releases/download/v0.3.1/convkit-installer.sh | sh
Video file too large? --max-size makes it fit.conv clip.mp4 --max-size 5mb
writes clip-5mb.mp4 at or under 5 MB, for upload limits and attachments.
conv chooses the resolution, frame rate and bitrates together, so a tight
target trims a little from each rather than everything from one. It encodes
in two passes, measures the result, and plans again if it came out over (three
attempts at most). Targets: mp4, mov, mkv and webm. A target too small to look
good is converted only after a warning, a suggested size and a [y/N]
question.
Video knobs.--fps caps the frame rate and --crf sets the encoder's
quality anchor on video and GIF targets. --resize now applies to video and
GIF as well as images. A cap the source already satisfies changes nothing,
and a stream copy stays a stream copy.
--upscale.--resize never enlarges, on any target. --upscale lets it,
with a warning that enlarging adds no detail and makes a larger file. Above
four times the source's pixels, conv also shows the output size and a rough
file-size estimate, and asks [y/N].
--yes now answers these questions as well as the backend-install prompt.
Without a terminal to ask and without --yes, nothing is converted and conv
exits 2 (confirmation_required in --json).
Changed
Images no longer enlarge without --upscale. Until 0.3.0, --resize 1600x900 on a 320x240 image produced 1200x900; it now leaves the image at
320x240. Add --upscale for the old result.
conv capabilities --json:defaults moved from the top level into each targets[] row, because a default belongs to a pair (mkv → mp4 and mkv → webm
carry different CRFs). A key with no default is left out rather than null.
GIF: an explicit --fps or --resize replaces the GIF defaults (15 fps,
640 px wide), capped at the source, even when the source already satisfies it.
Fixed
conv capabilities listed quality 92 as a default for video formats, which
refuse --quality. Its "tuning flags when writing X" line also claimed flags
that only some sources into X take.
conv --help printed internal developer notes under several flags.
Install convkit 0.3.0
Install prebuilt binaries via shell script
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/shdwfruit/convkit/releases/download/v0.3.0/convkit-installer.sh | sh
Not a convkit release. This pre-release holds third-party files that convkit's CI downloads, so that a CI run does not depend on an outside mirror being up. To install convkit, use the latest release.
