Generate

Commands that make pictures, video takes, voice lines and words, list models, save finished jobs, and run canvas boards.

Every command on this page needs a signed-in computer. Signed out, each one prints You are not signed in. Ask your agent to run minicut login. and exits with code 10. Flags every command takes, and every exit code, are on Global options.

minicut gen image

Makes one picture for a character, location, prop or style in your library.

Usage

minicut gen image OUT_FILE --prompt-file PROMPT_FILE --model MODEL --target KIND:ID [--project PROJECT] [--episode EPISODE] [--size W*H | --res RESOLUTION --aspect RATIO | --quality QUALITY] [--on {blank|PICTURE}] [--ref PICTURE ...] [--name NAME] [--json]

Replace the following:

• OUT_FILE: the picture file to make, such as lin-xia.png.
• PROMPT_FILE: a text file with the words that describe the picture.
• MODEL: an image model name from minicut models.
• KIND:ID: what the picture is for. KIND is character, location, prop or style, and ID is the item's id.
• PROJECT: a project id or the path to a project folder.
• EPISODE: an episode id such as ep01.
• W*H: the size in pixels, such as 1024*1536.
• RESOLUTION, RATIO, QUALITY: values the model lists in its settings.
• PICTURE: a picture file on your computer.
• NAME: the short name people read on the asset card.

Details

This command uses credits. It shows what it will do and asks before it spends; --yes skips the question. It does not wait for the picture. It sends the job and returns, and minicut jobs saves the picture into your library when it is done.

With no size flag, the picture is asked at the project's screen shape when the model lists it. When it does not, the command says so and asks you to add --aspect.

Options

Examples

Find an image model

minicut models --kind image

Make a character picture

minicut gen image lin-xia.png --prompt-file lin-xia.txt --model MODEL --target character:lin-xia --project rain-shop

Replace MODEL with a name from the list. This sends nothing until you say yes.

Output

Without --yes, the command says Nothing was sent. and prints the line to run after your yes. With --json it prints "state": "needs_yes" and that line as next_step. With the --yes line, it sends the job and prints its job id.

See also

minicut jobs, minicut models, minicut save

minicut gen video

Makes one video take for a shot, from the shot's own folder.

Usage

minicut gen video CLIP_FOLDER [--again] [--reason REASON] [--name NAME] [--json]

Replace the following:

• CLIP_FOLDER: the shot's folder, such as episodes/ep01/clips/c03.
• REASON: a few words on why you make this take.
• NAME: the short name people read on the take.

Details

The command reads the shot's settings, prompt and pictures from its folder. This command uses credits. It shows what it will do and asks before it spends; --yes skips the question. It sends the job and returns, and minicut jobs saves the take into the shot's folder.

One shot has one job at a time. If the shot already has a job queued or running, the command stops and names that job. Stop it first, or add --again to make another take beside it.

If you make a video take before the Mock loop is approved, the command prints a warning and records it. It does not refuse. See Checks and loops.

Options

Examples

Make a take for shot 3

minicut gen video episodes/ep01/clips/c03

Output

The same as minicut gen image: the --yes line first, then the job id after your yes.

Errors

See also

minicut jobs, minicut check take

minicut gen voice

Makes one spoken line for the mock, saved in the shot's folder.

Usage

minicut gen voice OUT_FILE --text-file TEXT_FILE --voice VOICE --target line:CLIP-N --project PROJECT [--episode EPISODE] [--model MODEL] [--json]

Replace the following:

• OUT_FILE: the sound file to make, such as l01.mp3.
• TEXT_FILE: a text file with the words to speak.
• VOICE: a voice name. minicut voices lists the voices a voice model takes.
• CLIP-N: the shot and the line number, such as c03-1.
• PROJECT: a project id or the path to a project folder.
• EPISODE: an episode id such as ep01.
• MODEL: a voice model name from minicut models.

Details

This command uses credits. It shows what it will do and asks before it spends; --yes skips the question. The finished line is saved in the shot's folder, under lines/.

Options

Examples

Make line 1 in shot 3

minicut gen voice l01.mp3 --text-file line.txt --voice VOICE --target line:c03-1 --project rain-shop --episode ep01

Replace VOICE with a voice name.

See also

minicut gen batch, minicut jobs

minicut gen MODEL

Runs any other model by its name: music, sound split, lip sync, upscale, clip repair and 3D.

Usage

minicut gen MODEL OUT_FILE [--prompt PROMPT | --prompt-file PROMPT_FILE] [--set SETTING=VALUE ...] [--in INPUT=FILE ...] --target KIND:ID [--project PROJECT] [--episode EPISODE] [--name NAME] [--json]

Replace the following:

• MODEL: a model name from minicut models.
• OUT_FILE: the file to make.
• PROMPT: the words for the model, in quotes.
• PROMPT_FILE: a text file with the words for the model.
• SETTING=VALUE: one setting the model lists, such as seconds=8. Repeat it for more.
• INPUT=FILE: one input file the model takes. Repeat it for more.
• KIND:ID: what the file is for. KIND is character, location, prop, style or canvas.

Details

minicut models MODEL lists the settings and inputs a model takes. This command uses credits. It shows what it will do and asks before it spends; --yes skips the question.

Options

See also

minicut models, minicut jobs

minicut gen batch

Sends many gen commands and asks once for all of them.

Usage

minicut gen batch LIST_FILE [--no-wait] [--json]

Replace the following:

• LIST_FILE: a text file with one gen command a line, written without minicut gen in front.

Details

Each line is image ..., voice ... or video CLIP_FOLDER, with its flags. Leave --yes out of every line. This command uses credits. It shows what it will do and asks before it spends; --yes skips the question.

The --yes line sends the items one at a time. When your account has no free slot, it waits up to 9 minutes for one, and nothing is sent while it waits. Running the same --yes line again never sends an item twice. A shot can appear on one line only, unless that line has --again.

Options

Examples

Make three pictures and two takes at once

image lin-xia.png --prompt-file lin-xia.txt --model MODEL --target character:lin-xia --project rain-shop
image shop.png --prompt-file shop.txt --model MODEL --target location:shop --project rain-shop
image umbrella.png --prompt-file umbrella.txt --model MODEL --target prop:umbrella --project rain-shop
video episodes/ep01/clips/c01
video episodes/ep01/clips/c02

minicut gen batch shots.txt

Output

Without --yes, one line per item, then Nothing was sent. and the --yes line. With the --yes line, one result per item, then how many of them were sent.

See also

minicut jobs

Lists your open jobs and saves the finished ones into your project.

Usage

minicut jobs [--wait SECONDS] [--json]
minicut jobs --get JOB_ID [--json]
minicut jobs --stop JOB_ID [--json]

Replace the following:

• SECONDS: how long to keep checking, in seconds.
• JOB_ID: the id of one job, as minicut jobs lists it.

Details

minicut gen and minicut board run send jobs and do not wait for them. minicut jobs reads every open job on your account. It saves each finished result into the project or board it was made for, and lists the rest by id. A job made for a project that is not on this computer is listed and left open.

--wait keeps checking until every job it can save is saved, or the time runs out. --get saves one job here, even when it was already saved. --stop stops one job.

Options

Examples

Save finished jobs, waiting up to 5 minutes

minicut jobs --wait 300

Output

One line per job: job <id>: saved <files>, or its state, such as running. A failed job prints failed, its error code, and a Why: line with the reason. Exit codes: 0 when nothing failed, 6 when --wait ran out with a job still running, 7 when minicut.ai did not answer. A failed job exits with its own code, listed on Global options.

See also

minicut gen image, minicut shots

minicut models

Lists the models you can use, or shows one model's settings.

Usage

minicut models [--kind KIND] [--search WORDS] [--json]
minicut models MODEL [--json]

Replace the following:

• KIND: one of image, video, voice, audio, music, sfx, lipsync, upscale, videofx, audiofx or mesh3d.
• WORDS: words to search model names for.
• MODEL: one model's name.

Details

The list comes from minicut.ai and is saved on this computer. Offline, the command prints the saved list with its age. --kind audio lists the voice models and every music and sound model.

Options

Examples

List the video models

minicut models --kind video

Output

One line per model: its name, kind and status. minicut models MODEL prints that model's settings and page. Exit codes: 0 when it listed the models, 7 when it printed the saved list because minicut.ai did not answer.

See also

minicut write

Writes words for a text box with a writing model.

Usage

minicut write TEXT [--model MODEL] [--json]

Replace the following:

• TEXT: what to write, in quotes.
• MODEL: a writing model name.

Details

This command uses credits. It shows what it will do and asks before it spends; --yes skips the question. With no --model, it uses the default writing model. A name the account does not have is refused, with the names it does have.

Options

Examples

Ask for a one-line title

minicut write "A one-line title for a rainy night shop drama"

Output

After your yes, the command prints the words. Exit code 3 means it is waiting for your yes.

See also

minicut board

Makes, edits, checks and runs a canvas board, which is a file you edit.

Usage

minicut board new [--project PROJECT --episode EPISODE] [--title TITLE] [--json]
minicut board check FILE [--json]
minicut board run FILE {--box BOX ... | --all} [--wait SECONDS] [--json]
minicut board add FILE --kind KIND [--id BOX] [--x N --y N] [--label LABEL] [--prompt PROMPT] [--model MODEL] [--parent BOX]
minicut board move FILE --box BOX --x N --y N
minicut board set FILE --box BOX [--label LABEL] [--prompt PROMPT] [--text TEXT] [--model MODEL] [--seconds N] [--file PATH] [--fields JSON]
minicut board {connect|disconnect} FILE --from BOX --to BOX
minicut board delete FILE --box BOX [--box BOX ...]
minicut board view FILE --view {workflow|storyboard}
minicut board retype FILE --box BOX --kind {character|location|prop}

Replace the following:

• FILE: the board file.
• PROJECT, EPISODE: the project and episode the board belongs to.
• TITLE: the board's title.
• BOX: a box id on the board.
• KIND: the kind of box, such as image, video or character.
• N: a number: a position, or a length in seconds.
• PATH: a file in the board's folder.
• JSON: several fields at once, as JSON.

Details

The board file holds the boxes and the wires between them. Each edit command changes the file the same way the open 画布 tab in the MiniCut app does, and the open tab redraws by itself.

• new makes an empty board. With --project and --episode it goes in the episode's canvas folder. Without them it goes in its own folder in your MiniCut home folder.
• check reads the board for problems. It finds boxes with the same id, wires to missing boxes, wires that are not allowed, and missing upload files.
• run makes the listed boxes. It uses credits. It shows what it will do and asks once for all the boxes; --yes skips the question. A box that is refused keeps its reason on the board, and the other boxes still go. --wait saves the results into the board's folder.

Options

Examples

Make a board for episode 1

minicut board new --project rain-shop --episode ep01 --title "Opening"

Make every box on a board

minicut board run episodes/ep01/canvas/opening.board.json --all

Output

check prints Board ok: with the number of boxes and wires, or one line per problem. Each edit prints Board <command>: ok with the board's new revision number, and the new box id after add.

See also

minicut jobs, minicut write