@thenavidm/midjourney-mcp-cli 1.0.0 → 1.2.0

package/README.md

@@ -1,35 +1,23 @@
-<div align="center">
-  <img src="https://cdn.navid.media/connectors/midjourney-icon.png" alt="Midjourney" width="88">
-</div>
+<img src="https://cdn.navid.media/connectors/midjourney-icon-solid.png" alt="Midjourney" width="88">
 
 # Midjourney MCP + CLI
 
+[![npm](https://img.shields.io/npm/v/@thenavidm/midjourney-mcp-cli?color=orange&label=npm)](https://www.npmjs.com/package/@thenavidm/midjourney-mcp-cli)
 [![Licence](https://img.shields.io/badge/licence-MIT-green)](./LICENSE)
 [![YouTube](https://img.shields.io/badge/YouTube-@thenavidm-red?logo=youtube&logoColor=white)](https://youtube.com/@thenavidm?sub_confirmation=1)
 [![X](https://img.shields.io/badge/X-@thenavidm-black?logo=x)](https://x.com/thenavidm)
 
-Midjourney MCP server and CLI for Claude Code and AI agents. 27 tools for generating images, following jobs to completion, downloading the real files, and building moodboards that make a style reusable.
+Midjourney MCP server and CLI for Claude Code and AI agents. 32 tools for generating images, following jobs to completion, downloading the real files, and building moodboards that make a style reusable.
 
 Midjourney publishes no API, so this drives a real Chrome that is signed in as you.
 
 There is no key to paste and no cookie to export. You sign in once, in a window, and the session lives in a browser profile rather than in a config file.
 
-27 tools, on both surfaces. It waits for jobs to finish and hands back the actual files, not a screenshot of them.
+32 tools, on both surfaces. It waits for jobs to finish and hands back the actual files, not a screenshot of them.
 
 Built and maintained by [Navid Moazzez](https://navid.me).
 
-```
-You: make a moodboard for cold Nordic product shots, fill it, then shoot a jar of face cream in that style
-
-Claude: Built the board and used it.
-
-  moodboard  Nordic Skincare | Still Life        4 images
-  job        f44b0a9d-e184-4e18-811f-a1ef9482d286
-  prompt     a ceramic jar of face cream, lid beside it
-  style      from the moodboard, not the prompt
-
-  ~/Downloads/midjourney/f44b0a9d-0.png   1.8 MB   960x1200
-```
+<img src="https://cdn.navid.media/repos/midjourney-mcp-cli.gif?v=2" alt="Claude Code using the Midjourney MCP server" width="520">
 
 ## Two ways to use it
 
@@ -73,12 +61,11 @@ Every other client is in [section 3](#3-install).
 
 ### Which one
 
-| What you are doing | Use |
+| Where you are | What you can reach |
 |---|---|
-| Inside a conversation with an agent | MCP |
-| On claude.ai or your phone | Neither. The browser is on your machine, so a cloud connector cannot reach it |
-| Piping, scripting, cron, CI | CLI |
-| A one-off question in a terminal | CLI |
+| An agent that can run shell commands, like Claude Code or Cursor | Both. The CLI is the cheaper one: it costs nothing until you type it |
+| claude.ai or a phone | Neither. Your logged-in browser is on this machine, and a cloud connector cannot reach it |
+| A terminal, a script, cron or CI | The CLI only. There is no MCP client in a shell |
 
 They are the same program reading the same tool definitions, so anything one
 can do, the other can.
@@ -131,8 +118,8 @@ All 27 with their arguments are in [section 6](#6-tools).
 | 11 | [Your data](#11-your-data) | What is stored and where |
 | 12 | [Risks](#12-risks) | Read this before you install |
 | 13 | [Troubleshooting](#13-troubleshooting) | When something breaks |
-| | [Environment variables](#environment-variables) | Every knob, and its default |
-| 14 | [FAQ](#14-faq) | Including what an MCP server is |
+| 14 | [Environment variables](#14-environment-variables) | Every knob, and its default |
+| 15 | [FAQ](#15-faq) | Including what an MCP server is |
 
 ## 1. What you can ask it
 
@@ -141,6 +128,7 @@ All 27 with their arguments are in [section 6](#6-tools).
 - Make a moodboard for cold Nordic product shots, fill it, then shoot a jar of face cream in that style
 - Generate four logo concepts at low stylize so they stay literal, and save them
 - Vary the second one, strong, and save the results
+- Upscale that one and turn it into a video
 - Take that last image's seed and try it again with chaos 40
 - What is in my Midjourney queue right now?
 - Download everything I generated today into ./renders
@@ -180,8 +168,6 @@ Node 22 or newer, and Google Chrome. Nothing else.
 
     npx -y @thenavidm/midjourney-mcp-cli --version
 
-> [!NOTE]
-> Not published to npm yet. Until it is, clone the repo and run `npm install && npm run build`, then use `node dist/index.js` wherever this page says `npx -y @thenavidm/midjourney-mcp-cli@latest`.
 
 Node 22 is the floor because the browser connection uses the global `WebSocket`
 that landed in that release. That is also why it has no dependency doing it.
@@ -301,7 +287,7 @@ authorises a charge.
 
 An MCP server is expensive and a CLI is free.
 
-The `tools/list` payload for these 27 tools is about **8,900 tokens**, plus the
+The `tools/list` payload for these 32 tools is about **11,062 tokens**, plus the
 server instructions. That is charged on every turn of every conversation, used
 or not, because the descriptions are long and carry the parameter grammar.
 
@@ -310,11 +296,10 @@ the model pays only when it runs something.
 
 So the two are not competing:
 
-| Where the work happens | Surface |
+| Where you are | What you can reach |
 |---|---|
-| Inside a conversation with an agent | MCP |
-| Piping, scripting, cron, CI | CLI |
-| A one-off question in a terminal | CLI |
+| An agent that can run shell commands | Both. The CLI costs nothing until you type it |
+| A terminal, a script, cron or CI | The CLI only |
 
 ## 6. Tools
 
@@ -326,6 +311,11 @@ So the two are not competing:
 | `submit_imagine` | Submit and return the job id without waiting. Spends |
 | `rerun_job` | Run an existing job again, optionally with new wording or at HD. Spends |
 | `vary_image` | Four variations of one image from a grid, subtle or strong. Spends |
+| `upscale_image` | Upscale one image to full resolution, subtle or creative. Spends |
+| `animate_image` | Turn one image into a video. Spends |
+| `pan_image` | Extend the frame left, right, up or down. Spends |
+| `zoom_out` | Pull the camera back and fill the new space. Spends |
+| `remix_image` | Re-render an image against a new prompt, keeping its composition. Spends |
 | `submit_raw_job` | Send a job type this server does not model yet. Spends |
 
 ### Following work
@@ -429,29 +419,42 @@ At the terminal, Midjourney's own spellings work as aliases: `--ar`, `--sref`,
 
 ## 9. Moodboards
 
-A moodboard is a curated pile of reference images. Naming one is far more
-reliable than describing a look in words, because the board *is* the look.
+A moodboard is a set of images the account has curated. Applying one is the same
+mechanism the web app's Personalize panel uses: the board becomes a
+personalization code on the prompt, resolved for you from its name.
+
+```bash
+midjourney-cli imagine "a ceramic jar of face cream" --moodboard "Nordic Skincare" --confirm
+```
+
+That is worth stating because there is a plausible wrong way to do the same
+thing: sampling the board's images into `--sref` URLs. It only registers at a
+high `--sw`, and that weight is what makes output look over-processed. The
+personalization code needs no weight and does not compete with your wording.
 
 The loop:
 
 ```bash
-midjourney-cli create-moodboard "Nordic Skincare | Still Life"
+midjourney-cli create-moodboard "Nordic Skincare"
 midjourney-cli imagine "<a long, specific style description>" --confirm
 midjourney-cli add-to-moodboard "Nordic Skincare" --job-id <job>
-midjourney-cli imagine "a ceramic jar of face cream, lid beside it" \
-  --moodboard "Nordic Skincare" --sw 400 --confirm
+midjourney-cli imagine "a ceramic jar of face cream" --moodboard "Nordic Skincare" --confirm
 ```
 
-After the third line the style is a name, and a nine-word prompt reproduces it.
-
 Partial names work: `"High Fashion"` finds `"High Fashion | Woman"`. An ambiguous
 name errors with the candidates rather than guessing, because picking the wrong
-board costs a generation to discover. References are sampled across the board
-rather than taken from the front, so a 242-image board does not always draw on
-its oldest images.
+board costs a generation to discover.
+
+`--p` accepts several codes, so a moodboard and a personalization profile apply
+together. `profile` is the other kind of personalization: it biases toward
+images the account has rated, rather than toward a set of pictures.
+
+### Leave the sliders alone
 
-`profile` does something different: it biases toward images the account has
-rated, rather than toward a set of pictures.
+Midjourney's own defaults for stylize, weirdness and variety sit near the
+minimum. Raising `stylize` trades fidelity for a prettier, more generic image.
+Set them when you want that; otherwise a strong result comes from the prompt and
+the reference.
 
 ## 10. How it works
 
@@ -524,7 +527,7 @@ Start with `doctor`. It orders the checks so the first failure is the one to fix
 | Every command times out at once | A native dialog was left open in the window. Dialogs are auto-dismissed now; if it persists, close the tab |
 | Downloads are empty or fail | The asset URL expired. Re-read the job with `get_job` for fresh URLs |
 
-## Environment variables
+## 14. Environment variables
 
 Every one of these is optional. The defaults are what you want unless you are doing something unusual.
 
@@ -553,7 +556,7 @@ Every one of these is optional. The defaults are what you want unless you are do
 | `MIDJOURNEY_HTTP_HOST` | `127.0.0.1` | Interface for `--http` |
 | `MIDJOURNEY_HTTP_TOKEN` | unset | Bearer token. Required to listen off loopback |
 
-## 14. FAQ
+## 15. FAQ
 
 <details>
 <summary><b>What is an MCP server?</b></summary>
@@ -650,7 +653,7 @@ Navid Moazzez is a leading AI business strategist, and the host of the AI Creato
 **Links**
 
 - Personal website: [navid.me](https://navid.me)
-- Store: [navid.bio](https://navid.bio)
+- Link in bio: [navid.bio](https://navid.bio)
 - Navid Media: [navid.media](https://navid.media)
 - YouTube: [@thenavidm](https://youtube.com/@thenavidm?sub_confirmation=1) and [@thenavidai](https://youtube.com/@thenavidai?sub_confirmation=1)
 - X: [@thenavidm](https://x.com/thenavidm)

package/SKILL.md

@@ -68,10 +68,10 @@ The ones worth knowing:
 | Argument | What it does |
 |---|---|
 | `aspect` | `"16:9"`, `"3:2"`, `"1:1"` |
-| `stylize` | 0-1000. Low follows the prompt, high looks prettier and drifts |
-| `chaos` | 0-100. How different the four results are from each other |
+| `stylize` | 0-1000. Leave unset: the default is low and raising it costs fidelity |
+| `chaos` | 0-100. Leave unset unless exploring |
 | `seed` | Reuse with an identical prompt to iterate on one image, not roll a new one |
-| `style_refs` | Style references: an image URL, a numeric code, or `"random"` |
+| `style_refs` | One specific reference. For a curated look use `moodboard` instead |
 | `omni_refs` | Carry a character or object across images. The v7 replacement for `--cref` |
 | `image_prompts` | Direct image URLs, used as visual input |
 | `negative` | Things to keep out, e.g. `"text, watermark"` |
@@ -82,52 +82,117 @@ The ones worth knowing:
 At the terminal, Midjourney's own spellings work as aliases: `--ar`, `--sref`,
 `--oref`, `--iw`, `--sw`, `--ow`, `--q`, `--no`, `--v`.
 
-## Moodboards are the best styling tool here
+## Use the newest model unless told otherwise
 
-The account has curated boards of reference images. Naming one is far more
-reliable than describing a look in words, because the board *is* the look.
+The default is the current model, v8.2. Only pin an older one when the user asks
+for it, or when they are iterating on an image made with it and want the match.
 
-```
-imagine(prompt: "a model in an ivory suit on a coastal cliff",
-        moodboard: "High Fashion", moodboard_refs: 4, confirm: true)
+## Moodboards are the strongest styling tool, and they are not srefs
+
+A moodboard is a collection of images the account has curated. Applying one is
+the same mechanism as selecting it in the web app's Personalize panel: the board
+becomes a **personalization code** on the prompt.
+
+```bash
+midjourney-cli imagine "a ceramic jar of face cream" --moodboard "Nordic Skincare" --confirm
 ```
 
-Partial names work. An ambiguous name errors with the candidates rather than
-guessing, because picking the wrong board costs a generation to find out.
+The code is the board's id with an `m` in front, and the tool resolves it for
+you. This matters because there is an obvious wrong way to do the same thing:
+sampling the board's images into `--sref` URLs. That approximation only shows up
+at a high `--sw`, and that weight is what makes output look processed. Use the
+moodboard.
 
-`list_moodboards` shows them with image counts. A board showing 0 images is
-empty and cannot be referenced yet. `get_moodboard` shows exactly which
-references a generation would use.
+`--p` accepts more than one code, so a moodboard and a personalization profile
+can apply together.
 
-`profile` does something different: it biases toward images the account has
-rated, rather than toward a set of pictures. `list_personalized_profiles`
-reports how many ratings each is built on, and one with a low count barely
-moves the result.
+Three styling tools, three jobs:
 
-## Building a moodboard, which is the real workflow
+| Want | Use |
+|---|---|
+| A look the account has curated | `moodboard` |
+| One specific reference image or code | `style_refs` |
+| The same character or object across images | `omni_refs` |
+| The account's learned taste | `profile` |
 
-Generate a style, keep what works, reuse it. That loop is what the boards are
-for, and it is worth driving deliberately.
+## Leave stylize, chaos and weird alone
 
-```
-create_moodboard(title: "Nordic Skincare | Still Life")
-imagine(prompt: "<a long, specific style description>", confirm: true)
-add_to_moodboard(moodboard: "Nordic Skincare", job_id: "<the job>", confirm: true)
-```
+Midjourney's own defaults sit low, and the sliders in its settings panel start
+near the minimum. Raising `stylize` trades fidelity for a prettier, more
+generic image; `chaos` and `weird` are for exploring, not for quality.
 
-After that the style is a name. A nine-word prompt reproduces it:
+Reach for them only when the user asks for more stylisation or more variety. A
+strong result comes from the prompt and the reference, not from the knobs.
 
+## Write prompts that leave nothing to chance
+
+Vague adjectives produce generic images. Specificity is what earns the output.
+Compare:
+
+> a beautiful woman in a knit set, studio, high quality, 8k
+
+against
+
+> a woman in three-quarter turn against a seamless clay cyclorama, rib-knit set
+> in oat, one hand at the hip, chin level, lit by a single large softbox high
+> and slightly left with a white bounce filling the shadow side so the falloff
+> is gentle, shot on medium format at f5.6, skin unretouched with visible pores,
+> nothing else in frame
+
+The second names the pose, the light source and its position, the fill, how the
+falloff should behave, the camera and aperture, what the skin keeps, and what
+must not appear. Every one of those is a decision the model would otherwise make
+for you.
+
+Things worth stating explicitly, because the model will invent them otherwise:
+
+- **Where the light comes from**, and what fills the shadow
+- **What the camera is**, and the aperture, which sets how much falls off
+- **What the subject is doing** with hands, shoulders, gaze
+- **What must not be in frame**, via the prompt or `negative`
+- **The palette**, named as colours rather than a mood
+- **Skin, fabric and surface texture**, or you get plastic
+
+`raw` is worth setting for anything photographic: it applies less of
+Midjourney's automatic prettification.
+
+## The pipeline is where the good work happens
+
+One generation is a draft. The tools exist to take it somewhere:
+
+```bash
+# 1. Explore cheaply. Draft mode is fast and costs less.
+midjourney-cli imagine "<a long, specific prompt>" --draft --confirm
+
+# 2. Once the composition is right, run it properly and keep the seed.
+midjourney-cli imagine "<the same prompt>" --seed 501058481 --raw --confirm
+
+# 3. Push the one that is closest.
+midjourney-cli vary-image <job> --index 2 --confirm          # subtle, or --strong
+midjourney-cli remix-image <job> --index 2 --prompt "<reworded>" --confirm
+
+# 4. Give it room, or take the frame wider.
+midjourney-cli zoom-out <job> --index 2 --zoom-factor 150 --confirm
+midjourney-cli pan-image <job> --index 2 --direction left --confirm
+
+# 5. Finish it.
+midjourney-cli upscale-image <job> --index 2 --confirm       # subtle, or creative
+midjourney-cli animate-image <job> --index 2 --motion "slow push in" --confirm
+midjourney-cli download-job <job> --out-dir ./renders
 ```
-imagine(prompt: "a ceramic jar of face cream, lid beside it",
-        moodboard: "Nordic Skincare", moodboard_refs: 4, confirm: true)
-```
 
-Push `style_weight` up (300-500) when the look should dominate the prompt, and
-down when the subject matters more than the styling.
+`seed` is how you iterate rather than reroll: the same prompt and seed gives a
+near-identical image, so a small wording change shows its own effect instead of
+a new roll of the dice.
+
+`zoom_out` before adding a headline or a logo: it gives the composition room
+without regenerating it.
+
+`upscale` subtle keeps what is there; creative reworks detail as it enlarges, so
+it is the wrong choice when the image is already right.
 
-`add_to_moodboard` takes a whole job at once, or specific `indexes`, or bare
-`urls`. `remove_from_moodboard` is how a board stays sharp, and it cannot be
-undone, so it asks for confirmation.
+`animate_image` takes a `motion` note, not a scene: it is appended to the
+image's own prompt, because Midjourney reads that field as what is in the frame.
 
 ## Working with results
 

package/dist/api/download.js

@@ -37,7 +37,8 @@ export function fileNameFor(url, jobId, index) {
     // job in a folder, so the job id goes in front. The stem itself is dropped
     // when it is just the grid position again: `<job>-0-0_0.png` says nothing
     // that `<job>-0.png` does not.
-    const redundant = /^\d+_\d+$/.test(stem);
+    // `0_2` for a grid image, `2` for a video segment: both repeat the index.
+    const redundant = /^\d+(_\d+)?$/.test(stem);
     const safeStem = redundant ? "" : stem.replace(/[^a-zA-Z0-9._-]/g, "-").slice(0, 60);
     return `${jobId}-${index}${safeStem ? `-${safeStem}` : ""}${extension}`;
 }

package/dist/api/download.js.map

@@ -1 +1 @@
-{"version":3,"file":"download.js","sourceRoot":"","sources":["../../src/api/download.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;GAkBG;AAEH,OAAO,EAAE,KAAK,EAAE,SAAS,EAAE,MAAM,kBAAkB,CAAC;AACpD,OAAO,EAAE,QAAQ,EAAE,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AAG7D,OAAO,EAAE,eAAe,EAAE,aAAa,EAAE,eAAe,EAAE,MAAM,aAAa,CAAC;AAC9E,OAAO,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AACpC,OAAO,EAAE,WAAW,EAAE,MAAM,qBAAqB,CAAC;AASlD,qFAAqF;AACrF,MAAM,UAAU,WAAW,CAAC,GAAW,EAAE,KAAa,EAAE,KAAa;IACnE,IAAI,SAAS,GAAG,EAAE,CAAC;IACnB,IAAI,CAAC;QACH,SAAS,GAAG,QAAQ,CAAC,IAAI,GAAG,CAAC,GAAG,CAAC,CAAC,QAAQ,CAAC,CAAC;IAC9C,CAAC;IAAC,MAAM,CAAC;QACP,SAAS,GAAG,EAAE,CAAC;IACjB,CAAC;IAED,MAAM,SAAS,GAAG,OAAO,CAAC,SAAS,CAAC,IAAI,YAAY,CAAC,WAAW,CAAC,GAAG,CAAC,CAAC,CAAC;IACvE,MAAM,IAAI,GAAG,SAAS,CAAC,CAAC,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC,EAAE,SAAS,CAAC,MAAM,GAAG,OAAO,CAAC,SAAS,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;IAE/F,4EAA4E;IAC5E,2EAA2E;IAC3E,0EAA0E;IAC1E,+BAA+B;IAC/B,MAAM,SAAS,GAAG,WAAW,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;IACzC,MAAM,QAAQ,GAAG,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,OAAO,CAAC,kBAAkB,EAAE,GAAG,CAAC,CAAC,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;IACrF,OAAO,GAAG,KAAK,IAAI,KAAK,GAAG,QAAQ,CAAC,CAAC,CAAC,IAAI,QAAQ,EAAE,CAAC,CAAC,CAAC,EAAE,GAAG,SAAS,EAAE,CAAC;AAC1E,CAAC;AAED,SAAS,YAAY,CAAC,IAAY;IAChC,QAAQ,IAAI,EAAE,CAAC;QACb,KAAK,WAAW;YACd,OAAO,MAAM,CAAC;QAChB,KAAK,YAAY;YACf,OAAO,MAAM,CAAC;QAChB,KAAK,YAAY;YACf,OAAO,OAAO,CAAC;QACjB,KAAK,WAAW;YACd,OAAO,MAAM,CAAC;QAChB,KAAK,WAAW;YACd,OAAO,MAAM,CAAC;QAChB,KAAK,YAAY;YACf,OAAO,OAAO,CAAC;QACjB;YACE,OAAO,MAAM,CAAC;IAClB,CAAC;AACH,CAAC;AAED,uCAAuC;AACvC,MAAM,CAAC,KAAK,UAAU,OAAO,CAC3B,MAAwB,EACxB,GAAW,EACX,MAAc,EACd,QAAgB;IAEhB,IAAI,CAAC,eAAe,CAAC,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC;QAC/B,MAAM,IAAI,eAAe,CAAC,4BAA4B,GAAG,IAAI,EAAE,CAAC,EAAE,SAAS,CAAC,CAAC;IAC/E,CAAC;IAED,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,GAAG,MAAM,MAAM,CAAC,SAAS,CAAC,WAAW,CAAC,GAAG,CAAC,CAAC;IACrE,MAAM,MAAM,GAAG,MAAM,CAAC,IAAI,CAAC,MAAM,EAAE,QAAQ,CAAC,CAAC;IAE7C,IAAI,MAAM,CAAC,UAAU,KAAK,CAAC,EAAE,CAAC;QAC5B,MAAM,IAAI,eAAe,CACvB,uBAAuB,GAAG,gGAAgG,EAC1H,CAAC,EACD,YAAY,CACb,CAAC;IACJ,CAAC;IAED,MAAM,SAAS,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC;IAClC,MAAM,KAAK,CAAC,SAAS,EAAE,EAAE,SAAS,EAAE,IAAI,EAAE,CAAC,CAAC;IAC5C,MAAM,IAAI,GAAG,IAAI,CAAC,SAAS,EAAE,QAAQ,CAAC,CAAC;IACvC,MAAM,SAAS,CAAC,IAAI,EAAE,MAAM,CAAC,CAAC;IAE9B,OAAO,EAAE,GAAG,EAAE,IAAI,EAAE,KAAK,EAAE,MAAM,CAAC,UAAU,EAAE,SAAS,EAAE,QAAQ,EAAE,CAAC;AACtE,CAAC;AAQD,kEAAkE;AAClE,MAAM,CAAC,KAAK,UAAU,WAAW,CAC/B,MAAwB,EACxB,KAAa,EACb,UAA8B,EAAE;IAEhC,MAAM,GAAG,GAAG,MAAM,OAAO,CAAC,MAAM,EAAE,KAAK,CAAC,CAAC;IACzC,IAAI,CAAC,GAAG,EAAE,CAAC;QACT,MAAM,IAAI,aAAa,CACrB,UAAU,KAAK,2IAA2I,EAC1J,GAAG,EACH,YAAY,CACb,CAAC;IACJ,CAAC;IACD,IAAI,GAAG,CAAC,MAAM,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QAC5B,MAAM,IAAI,aAAa,CACrB,OAAO,KAAK,2CAA2C,GAAG,CAAC,MAAM,IAAI,GAAG,CAAC,SAAS,CAAC,CAAC,CAAC,KAAK,GAAG,CAAC,SAAS,GAAG,CAAC,CAAC,CAAC,EAAE,8DAA8D,EAC7K,GAAG,EACH,YAAY,CACb,CAAC;IACJ,CAAC;IAED,MAAM,MAAM,GACV,OAAO,CAAC,OAAO,IAAI,OAAO,CAAC,OAAO,CAAC,MAAM,GAAG,CAAC;QAC3C,CAAC,CAAC,OAAO,CAAC,OAAO;QACjB,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,CAAC;IAE1C,MAAM,MAAM,GAAG,OAAO,CAAC,MAAM,IAAI,MAAM,CAAC,MAAM,CAAC,WAAW,CAAC;IAC3D,MAAM,KAAK,GAAgB,EAAE,CAAC;IAC9B,MAAM,OAAO,GAAa,EAAE,CAAC;IAE7B,KAAK,MAAM,KAAK,IAAI,MAAM,EAAE,CAAC;QAC3B,MAAM,GAAG,GAAG,GAAG,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC;QAC9B,IAAI,CAAC,GAAG,EAAE,CAAC;YACT,OAAO,CAAC,IAAI,CAAC,SAAS,KAAK,kBAAkB,GAAG,CAAC,MAAM,CAAC,MAAM,WAAW,CAAC,CAAC;YAC3E,SAAS;QACX,CAAC;QACD,IAAI,CAAC;YACH,KAAK,CAAC,IAAI,CAAC,MAAM,OAAO,CAAC,MAAM,EAAE,GAAG,EAAE,MAAM,EAAE,WAAW,CAAC,GAAG,EAAE,GAAG,CAAC,EAAE,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC;QAClF,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YACf,OAAO,CAAC,IAAI,CAAC,SAAS,KAAK,KAAM,KAAe,CAAC,OAAO,EAAE,CAAC,CAAC;QAC9D,CAAC;IACH,CAAC;IAED,OAAO,EAAE,MAAM,EAAE,GAAG,CAAC,EAAE,EAAE,KAAK,EAAE,OAAO,EAAE,CAAC;AAC5C,CAAC"}
+{"version":3,"file":"download.js","sourceRoot":"","sources":["../../src/api/download.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;GAkBG;AAEH,OAAO,EAAE,KAAK,EAAE,SAAS,EAAE,MAAM,kBAAkB,CAAC;AACpD,OAAO,EAAE,QAAQ,EAAE,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AAG7D,OAAO,EAAE,eAAe,EAAE,aAAa,EAAE,eAAe,EAAE,MAAM,aAAa,CAAC;AAC9E,OAAO,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AACpC,OAAO,EAAE,WAAW,EAAE,MAAM,qBAAqB,CAAC;AASlD,qFAAqF;AACrF,MAAM,UAAU,WAAW,CAAC,GAAW,EAAE,KAAa,EAAE,KAAa;IACnE,IAAI,SAAS,GAAG,EAAE,CAAC;IACnB,IAAI,CAAC;QACH,SAAS,GAAG,QAAQ,CAAC,IAAI,GAAG,CAAC,GAAG,CAAC,CAAC,QAAQ,CAAC,CAAC;IAC9C,CAAC;IAAC,MAAM,CAAC;QACP,SAAS,GAAG,EAAE,CAAC;IACjB,CAAC;IAED,MAAM,SAAS,GAAG,OAAO,CAAC,SAAS,CAAC,IAAI,YAAY,CAAC,WAAW,CAAC,GAAG,CAAC,CAAC,CAAC;IACvE,MAAM,IAAI,GAAG,SAAS,CAAC,CAAC,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC,EAAE,SAAS,CAAC,MAAM,GAAG,OAAO,CAAC,SAAS,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;IAE/F,4EAA4E;IAC5E,2EAA2E;IAC3E,0EAA0E;IAC1E,+BAA+B;IAC/B,0EAA0E;IAC1E,MAAM,SAAS,GAAG,cAAc,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;IAC5C,MAAM,QAAQ,GAAG,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,OAAO,CAAC,kBAAkB,EAAE,GAAG,CAAC,CAAC,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;IACrF,OAAO,GAAG,KAAK,IAAI,KAAK,GAAG,QAAQ,CAAC,CAAC,CAAC,IAAI,QAAQ,EAAE,CAAC,CAAC,CAAC,EAAE,GAAG,SAAS,EAAE,CAAC;AAC1E,CAAC;AAED,SAAS,YAAY,CAAC,IAAY;IAChC,QAAQ,IAAI,EAAE,CAAC;QACb,KAAK,WAAW;YACd,OAAO,MAAM,CAAC;QAChB,KAAK,YAAY;YACf,OAAO,MAAM,CAAC;QAChB,KAAK,YAAY;YACf,OAAO,OAAO,CAAC;QACjB,KAAK,WAAW;YACd,OAAO,MAAM,CAAC;QAChB,KAAK,WAAW;YACd,OAAO,MAAM,CAAC;QAChB,KAAK,YAAY;YACf,OAAO,OAAO,CAAC;QACjB;YACE,OAAO,MAAM,CAAC;IAClB,CAAC;AACH,CAAC;AAED,uCAAuC;AACvC,MAAM,CAAC,KAAK,UAAU,OAAO,CAC3B,MAAwB,EACxB,GAAW,EACX,MAAc,EACd,QAAgB;IAEhB,IAAI,CAAC,eAAe,CAAC,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC;QAC/B,MAAM,IAAI,eAAe,CAAC,4BAA4B,GAAG,IAAI,EAAE,CAAC,EAAE,SAAS,CAAC,CAAC;IAC/E,CAAC;IAED,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,GAAG,MAAM,MAAM,CAAC,SAAS,CAAC,WAAW,CAAC,GAAG,CAAC,CAAC;IACrE,MAAM,MAAM,GAAG,MAAM,CAAC,IAAI,CAAC,MAAM,EAAE,QAAQ,CAAC,CAAC;IAE7C,IAAI,MAAM,CAAC,UAAU,KAAK,CAAC,EAAE,CAAC;QAC5B,MAAM,IAAI,eAAe,CACvB,uBAAuB,GAAG,gGAAgG,EAC1H,CAAC,EACD,YAAY,CACb,CAAC;IACJ,CAAC;IAED,MAAM,SAAS,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC;IAClC,MAAM,KAAK,CAAC,SAAS,EAAE,EAAE,SAAS,EAAE,IAAI,EAAE,CAAC,CAAC;IAC5C,MAAM,IAAI,GAAG,IAAI,CAAC,SAAS,EAAE,QAAQ,CAAC,CAAC;IACvC,MAAM,SAAS,CAAC,IAAI,EAAE,MAAM,CAAC,CAAC;IAE9B,OAAO,EAAE,GAAG,EAAE,IAAI,EAAE,KAAK,EAAE,MAAM,CAAC,UAAU,EAAE,SAAS,EAAE,QAAQ,EAAE,CAAC;AACtE,CAAC;AAQD,kEAAkE;AAClE,MAAM,CAAC,KAAK,UAAU,WAAW,CAC/B,MAAwB,EACxB,KAAa,EACb,UAA8B,EAAE;IAEhC,MAAM,GAAG,GAAG,MAAM,OAAO,CAAC,MAAM,EAAE,KAAK,CAAC,CAAC;IACzC,IAAI,CAAC,GAAG,EAAE,CAAC;QACT,MAAM,IAAI,aAAa,CACrB,UAAU,KAAK,2IAA2I,EAC1J,GAAG,EACH,YAAY,CACb,CAAC;IACJ,CAAC;IACD,IAAI,GAAG,CAAC,MAAM,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QAC5B,MAAM,IAAI,aAAa,CACrB,OAAO,KAAK,2CAA2C,GAAG,CAAC,MAAM,IAAI,GAAG,CAAC,SAAS,CAAC,CAAC,CAAC,KAAK,GAAG,CAAC,SAAS,GAAG,CAAC,CAAC,CAAC,EAAE,8DAA8D,EAC7K,GAAG,EACH,YAAY,CACb,CAAC;IACJ,CAAC;IAED,MAAM,MAAM,GACV,OAAO,CAAC,OAAO,IAAI,OAAO,CAAC,OAAO,CAAC,MAAM,GAAG,CAAC;QAC3C,CAAC,CAAC,OAAO,CAAC,OAAO;QACjB,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,CAAC;IAE1C,MAAM,MAAM,GAAG,OAAO,CAAC,MAAM,IAAI,MAAM,CAAC,MAAM,CAAC,WAAW,CAAC;IAC3D,MAAM,KAAK,GAAgB,EAAE,CAAC;IAC9B,MAAM,OAAO,GAAa,EAAE,CAAC;IAE7B,KAAK,MAAM,KAAK,IAAI,MAAM,EAAE,CAAC;QAC3B,MAAM,GAAG,GAAG,GAAG,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC;QAC9B,IAAI,CAAC,GAAG,EAAE,CAAC;YACT,OAAO,CAAC,IAAI,CAAC,SAAS,KAAK,kBAAkB,GAAG,CAAC,MAAM,CAAC,MAAM,WAAW,CAAC,CAAC;YAC3E,SAAS;QACX,CAAC;QACD,IAAI,CAAC;YACH,KAAK,CAAC,IAAI,CAAC,MAAM,OAAO,CAAC,MAAM,EAAE,GAAG,EAAE,MAAM,EAAE,WAAW,CAAC,GAAG,EAAE,GAAG,CAAC,EAAE,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC;QAClF,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YACf,OAAO,CAAC,IAAI,CAAC,SAAS,KAAK,KAAM,KAAe,CAAC,OAAO,EAAE,CAAC,CAAC;QAC9D,CAAC;IACH,CAAC;IAED,OAAO,EAAE,MAAM,EAAE,GAAG,CAAC,EAAE,EAAE,KAAK,EAAE,OAAO,EAAE,CAAC;AAC5C,CAAC"}

package/dist/api/jobs.d.ts

@@ -91,6 +91,70 @@ export declare function submitRaw(client: MidjourneyClient, jobType: string, ext
     jobIds: string[];
     raw: unknown;
 }>;
+/**
+ * Which upscaler a job can use, derived from the model that made it.
+ *
+ * The menu says "Subtle" and "Creative", but the wire wants a version-specific
+ * name: `v8r1_2x_subtle` for a v8.1 image, `v7_2x_creative` for a v7 one. Send
+ * the label and the request is refused. None of this is guessable, and it is
+ * why the first four attempts here were 400s.
+ */
+export declare function upscalerFor(version: string | undefined, style: "subtle" | "creative"): string;
+/** The model version a job was made with, read off the job itself. */
+export declare function versionOfJob(job: {
+    jobType?: string;
+    prompt?: string;
+} | undefined): string | undefined;
+/** Upscale one image from a grid. `subtle` keeps it; `creative` reworks detail. */
+export declare function submitUpscale(client: MidjourneyClient, jobId: string, index: number, style: "subtle" | "creative", options?: SubmitOptions & {
+    version?: string;
+}): Promise<{
+    jobIds: string[];
+    raw: unknown;
+}>;
+/**
+ * Animate one image into a video.
+ *
+ * Two things here are not guessable and both cost a rejected request to learn.
+ * The source is nested under `parentJob` as `image_num`, not the flat `index`
+ * every other job type uses. And `newPrompt` carries the image's own prompt,
+ * not a description of the motion: passing "slow push in" on its own is
+ * refused, because the field is the scene, not the movement.
+ *
+ * So a motion note is appended to the original prompt rather than replacing
+ * it, which is what the web app does when you type into its box.
+ */
+export declare function submitVideo(client: MidjourneyClient, jobId: string, index: number, opts?: SubmitOptions & {
+    motion?: string;
+    resolution?: "480" | "720";
+    mode?: "auto" | "manual";
+}): Promise<{
+    jobIds: string[];
+    raw: unknown;
+}>;
+/** Extend the frame in one direction. */
+export declare function submitPan(client: MidjourneyClient, jobId: string, index: number, direction: "left" | "right" | "up" | "down", opts?: SubmitOptions & {
+    prompt?: string;
+    fraction?: number;
+}): Promise<{
+    jobIds: string[];
+    raw: unknown;
+}>;
+/** Zoom out, filling the new space. 100 is no change; 200 doubles the frame. */
+export declare function submitZoomOut(client: MidjourneyClient, jobId: string, index: number, opts?: SubmitOptions & {
+    prompt?: string;
+    zoomFactor?: number;
+}): Promise<{
+    jobIds: string[];
+    raw: unknown;
+}>;
+/** Re-render one image against a new prompt, keeping its composition. */
+export declare function submitRemix(client: MidjourneyClient, jobId: string, index: number, prompt: string, opts?: SubmitOptions & {
+    strong?: boolean;
+}): Promise<{
+    jobIds: string[];
+    raw: unknown;
+}>;
 export type ListOptions = {
     limit?: number;
     cursor?: string;

package/dist/api/jobs.js

@@ -187,6 +187,130 @@ export async function submitRaw(client, jobType, extra, options = {}) {
     refreshView(client, options.refresh !== false);
     return { jobIds: extractJobIds(raw), raw };
 }
+/**
+ * The rest of the job types, read out of the web app's own bundle.
+ *
+ * Every shape below was extracted from the compiled client rather than guessed
+ * or clicked, so the field names are the app's own. That matters: `upscale`
+ * takes a `type`, `video` nests its source under `parentJob` with `image_num`
+ * rather than `index`, and `pan` carries a `fraction` and a `stitch` flag. None
+ * of that is inferable, and a wrong guess spends GPU time on a request that
+ * quietly does nothing.
+ */
+function baseBody(userId, options) {
+    return {
+        f: { mode: options.speed ?? "fast", private: options.private ?? false },
+        channelId: channelIdFor(userId),
+        metadata: {
+            isMobile: null,
+            imagePrompts: null,
+            imageReferences: null,
+            characterReferences: null,
+            depthReferences: null,
+            lightboxOpen: null,
+        },
+    };
+}
+async function submitShape(client, shape, options) {
+    const userId = await client.userId();
+    const body = { ...baseBody(userId, { speed: options.speed ?? client.config.defaultSpeed, ...options }), ...shape };
+    const raw = await client.request(ENDPOINTS.submit, { method: "POST", body, noRetry: true });
+    refreshView(client, options.refresh !== false);
+    return { jobIds: extractJobIds(raw), raw };
+}
+/**
+ * Which upscaler a job can use, derived from the model that made it.
+ *
+ * The menu says "Subtle" and "Creative", but the wire wants a version-specific
+ * name: `v8r1_2x_subtle` for a v8.1 image, `v7_2x_creative` for a v7 one. Send
+ * the label and the request is refused. None of this is guessable, and it is
+ * why the first four attempts here were 400s.
+ */
+export function upscalerFor(version, style) {
+    const v = (version ?? "8.1").toLowerCase().replace(/^v/, "").trim();
+    if (v.startsWith("8.2"))
+        return `v8r2_2x_${style}`;
+    if (v.startsWith("8"))
+        return `v8r1_2x_${style}`;
+    if (v.startsWith("niji 7") || v.startsWith("7"))
+        return `v7_2x_${style}`;
+    if (v.startsWith("6.1") || v.startsWith("niji 6"))
+        return `v6r1_2x_${style}`;
+    if (v.startsWith("6"))
+        return `v6_2x_${style}`;
+    // Older lines only ever had the fixed upscalers, with no subtle or creative.
+    return `v5_2x`;
+}
+/** The model version a job was made with, read off the job itself. */
+export function versionOfJob(job) {
+    const fromType = job?.jobType?.match(/^v(\d+)-(\d+)_/);
+    if (fromType)
+        return `${fromType[1]}.${fromType[2]}`;
+    const fromPrompt = job?.prompt?.match(/--v\s+(\S+)/);
+    if (fromPrompt)
+        return fromPrompt[1];
+    const niji = job?.prompt?.match(/--niji\s+(\S+)/);
+    if (niji)
+        return `niji ${niji[1]}`;
+    return undefined;
+}
+/** Upscale one image from a grid. `subtle` keeps it; `creative` reworks detail. */
+export async function submitUpscale(client, jobId, index, style, options = {}) {
+    const version = options.version ?? versionOfJob(await findJob(client, jobId));
+    return submitShape(client, { t: "upscale", type: upscalerFor(version, style), id: jobId.trim(), index }, options);
+}
+/**
+ * Animate one image into a video.
+ *
+ * Two things here are not guessable and both cost a rejected request to learn.
+ * The source is nested under `parentJob` as `image_num`, not the flat `index`
+ * every other job type uses. And `newPrompt` carries the image's own prompt,
+ * not a description of the motion: passing "slow push in" on its own is
+ * refused, because the field is the scene, not the movement.
+ *
+ * So a motion note is appended to the original prompt rather than replacing
+ * it, which is what the web app does when you type into its box.
+ */
+export async function submitVideo(client, jobId, index, opts = {}) {
+    const source = await findJob(client, jobId);
+    const original = (source?.prompt ?? "").trim();
+    const motion = opts.motion?.trim();
+    const newPrompt = motion ? `${original} ${motion}`.trim() : original;
+    return submitShape(client, {
+        t: "video",
+        videoType: `vid_1.1_i2v_${opts.resolution ?? "480"}`,
+        stitch: null,
+        newPrompt,
+        parentJob: { job_id: jobId.trim(), image_num: index },
+        animateMode: opts.mode ?? (motion ? "manual" : "auto"),
+    }, opts);
+}
+/** Extend the frame in one direction. */
+export function submitPan(client, jobId, index, direction, opts = {}) {
+    return submitShape(client, {
+        t: "pan",
+        newPrompt: opts.prompt ?? "",
+        direction,
+        fraction: opts.fraction ?? 0.5,
+        stitch: true,
+        id: jobId.trim(),
+        index,
+    }, opts);
+}
+/** Zoom out, filling the new space. 100 is no change; 200 doubles the frame. */
+export function submitZoomOut(client, jobId, index, opts = {}) {
+    return submitShape(client, {
+        t: "outpaint",
+        newPrompt: opts.prompt ?? "",
+        zoomFactor: opts.zoomFactor ?? 200,
+        id: jobId.trim(),
+        index,
+    }, opts);
+}
+/** Re-render one image against a new prompt, keeping its composition. */
+export function submitRemix(client, jobId, index, prompt, opts = {}) {
+    return submitShape(client, { t: "remix", strong: opts.strong ?? false, newPrompt: prompt, id: jobId.trim(), index }, opts);
+}
 /** Recent jobs for the signed-in account. */
 export async function listJobs(client, options = {}) {
     const userId = options.userId ?? (await client.userId());