Update to Kinetune

{
  "description": "Make lyric videos (9:16, 16:9, 1:1) and Spotify Canvas loops from a user's songs with Kinetune, through its MCP tools or the `kinetune` CLI. Use when the user wants a lyric video, a Spotify Canvas, a visualizer or promo video for a song, wants to upload a song or artist photos to Kinetune, browse Looks, check their credits, or mentions Kinetune.",
  "included_files": [
    {
      "relative_path": "references/cli.md",
      "size_in_bytes": 5257
    }
  ],
  "name": "kinetune",
  "skill_md_contents": "---\nname: kinetune\ndescription: Make lyric videos (9:16, 16:9, 1:1) and Spotify Canvas loops from a user's songs with Kinetune, through its MCP tools or the `kinetune` CLI. Use when the user wants a lyric video, a Spotify Canvas, a visualizer or promo video for a song, wants to upload a song or artist photos to Kinetune, browse Looks, check their credits, or mentions Kinetune.\n---\n\n# Kinetune\n\nKinetune turns a song into release-ready video:\n\n- **Lyric videos**: word-synced lyrics over a designed background with audio-reactive visuals, in 9:16 (vertical), 16:9 (wide) and 1:1 (square).\n- **Spotify Canvas**: a 5–8 second vertical loop designed from the song's cover, uploaded by the artist in Spotify for Artists.\n\nEvery video uses a **Look**, a complete reusable design. Reuse a saved Look, or have a new one designed from the song's cover art.\n\n## The one rule: quote, tell, wait for the go\n\nCreating a video spends the user's credits. Always:\n\n1. Quote the exact request (`quote_lyric_video` / `quote_canvas`, or `kinetune quote …`).\n2. Tell the user the credits, what they get, and their balance.\n3. Create only after they say yes, with the **same** input plus the `quote_id`. A quote is valid for 30 minutes.\n\nNever create without a quote and the user's agreement, and never pass the CLI's `--yes` on your own. If the user has already agreed to a budget, pass `--max-credits N` so the CLI refuses anything pricier.\n\n## Use the Kinetune tools\n\nWhen Kinetune is installed as a plugin or connector (ChatGPT, Claude, Cursor, VS Code…), its tools are already available: `get_account`, `list_songs`, `quote_canvas` and the rest. Use them and nothing else.\n\nApps that draw cards (ChatGPT, Claude and others with MCP Apps) show Kinetune's results as cards, so show rather than describe:\n\n- `show_looks` when the person wants to see or pick Looks; each plays as the real render. Pass `song_id` to preview them with their song. Their pick arrives as a message naming the Look.\n- `show_songs` when they want to see or pick a song.\n- A quote draws as an order ticket with a **Make it** button. Tapping it is their yes: the card creates the videos itself, so don't create them again.\n- `show_video` right after `create_lyric_video` or `create_canvas`, with the ids it returned. The card follows the render and offers the downloads, so you don't need `wait_for_video` then.\n\nIf the tools aren't available in a chat app, or anywhere Kinetune was added as a plugin or connector, the user hasn't connected it yet. Ask them to connect Kinetune and sign in with their Kinetune account; their app shows a Connect or Sign in step for it. Don't install anything, and don't switch to the CLI or the API instead. When unsure, ask them to connect.\n\nWithout a plugin or connector, two other ways reach the same operations:\n\n- **The CLI**, when the user works in a local coding agent with a terminal and installed only this skill, or asks for the CLI: `kinetune` (`npm install -g @kinetune/cli`, or `npx @kinetune/cli …` without installing). Check sign-in with `kinetune auth status`; if not signed in, ask the user to run `kinetune auth login` (it opens their browser) or set `KINETUNE_API_KEY`. Every command prints JSON when piped. See [references/cli.md](references/cli.md).\n- **REST** at `https://kinetune.com/api/v1` with `Authorization: Bearer <key>`, only when writing code for the user. The spec is at `https://kinetune.com/api/v1/openapi.json`.\n\nThey take the same inputs as the tools, so the steps below name each tool with its CLI command.\n\n## Make a video\n\n1. **Check the account**: `get_account` / `kinetune account`. Note the credit balance.\n2. **Find the song**: `list_songs` with `q` (part of the title or artist) / `kinetune songs list --q \"…\"`. Only songs with `analysis.status` `\"ready\"` can be used. Lists return 50 per page; page with `limit`/`offset` rather than listing everything.\n3. **No song yet?** Find or create the artist (`list_artists` / `create_artist`), then upload: `create_song` / `kinetune songs create --artist-id … --title … --audio ./song.wav --cover-art ./cover.jpg`. Audio is MP3, WAV or M4A up to 250 MB; the cover is a square JPEG or PNG, 1000–6000 px (3000×3000 recommended). Analysis (lyrics, beats, sections) takes about a minute: poll `get_song` until `analysis.status` is `\"ready\"`. If it fails, `reanalyze_song`. If the user has the song's lyrics, pass them too (see [The song's own lyrics](#the-songs-own-lyrics)).\n4. **Choose what to make** (below), then **quote** it and tell the user the credits.\n5. **Create** with the same input plus `quote_id`: `create_lyric_video` / `create_canvas`, or `kinetune create … --quote-id …`. It returns the video ids at once.\n6. **Wait**: `wait_for_video` (call again while `status` is `queued` or `processing`) / `kinetune videos wait VIDEO_ID`.\n7. **Deliver**: a completed video has one output per format with `download_url` (full quality), `web_url` (720p) and `poster_url`, signed for 7 days. Share the links, or `kinetune videos download VIDEO_ID --out ./videos`. A lyric video's poster is a frame with one whole lyric line on screen, made for the thumbnail where the user posts it; `poster_seconds` says where in the file it was cut, for platforms that take a cover frame offset instead of an image (TikTok).\n\n## The song's own lyrics\n\nSongs are transcribed to time every word, and a transcription can mishear one. When the user has the song's lyrics (pasted, Suno's lyrics, or an `.srt`, `.vtt` or `.lrc` file), give them to Kinetune: its videos then spell every word their way and break lines where they do, still timed from the audio.\n\n- With the upload: `lyrics` on `create_song` / `kinetune songs create … --lyrics-file ./song.txt`.\n- Later, or to fix them: `set_song_lyrics` / `kinetune songs set-lyrics SONG_ID --lyrics-file ./song.srt`. `clear_song_lyrics` / `kinetune songs clear-lyrics SONG_ID` goes back to the transcribed words.\n- Pass the text as the user has it: section tags like `[Chorus]` are fine, and a part sung again can be written once.\n- `get_song` → `analysis.lyrics.status`: `transcribed` (as heard), `checked` (theirs, timed from the audio; `corrected` and `added` say what changed), `waiting` (checked once the analysis finishes) or `mismatch`.\n- Lyrics that don't match what's sung, or cover only part of the song, are refused with the reason and the song stays as it was: tell the user what it says.\n- Videos made from then on use the lyrics. Videos already made stay as they are; making one again picks them up.\n\n## Choosing a lyric video\n\nRead `get_options` / `kinetune options` for the current categories, background sources, quality tiers, headlines and credit prices. Don't guess ids.\n\n- **Look**:\n  - `{\"mode\":\"existing\",\"id\":\"look_…\"}` reuses a saved Look exactly. It is the cheapest choice. Browse with `list_looks` (`scope`: `official`, `community`, `mine`; filter by `category`; pass the song's `artist_id` to leave out Looks that show another artist).\n  - `{\"mode\":\"new\",\"category\":\"…\",\"background\":{\"source\":\"ai-image\",\"quality\":\"high\"},\"direction\":\"…\",\"visibility\":\"private\"}` has a new Look designed from the cover.\n    - `direction` is the user's mood or references in plain words.\n    - The singer stays out of the background unless asked: `\"feature_artist\":\"always\"` shows them. It needs an AI background, artist photos (`add_artist_photos`), `\"display\":{\"cover\":false}` (the cover would hide them) and a private Look, which then only serves that artist's songs.\n- **Background sources** for a new Look: `ai-image`, `ai-video` (an AI image brought to life as a loop; costs more), `stock-photo`, `stock-video`. AI sources take a `quality` tier.\n- **`variations`** (1–4): several different New Looks, one video each.\n- **`aspect_ratios`**: any of `\"9:16\"`, `\"16:9\"`, `\"1:1\"`, each its own file. Each extra format adds credits.\n- **`display`**: which elements show (`title`, `artist`, `cover`, `lyrics`, `badges`, `headline`) and their sizes.\n- **`trim`**: render only part of the song (at least 8 seconds).\n\nAsk the user only what matters to them, usually the category or mood, the formats, and whether to reuse a Look. Default the rest.\n\n## Choosing a Canvas\n\n- **`source`**: `ai-video` (default, a video model animates an image made from the cover's world), `ai-image`, `stock-photo`, `stock-video`.\n- **`seconds`**: 5–8, default 8. Spotify loops it, so no part of the song is chosen.\n- **Optional**: `style` (`cinematic`, `dreamy`, `abstract`, `surreal`, `retro`, `minimal`, `dark`, `vibrant`, `nature`, `urban`), `direction`, `feature_artist` (`auto`, `always`, `never`; `always` needs artist photos), `quality` (`standard` or `high`), and `resolution` (`1080p` or `720p`).\n- **`variations`**: 1–4 different Canvases.\n\n## When something goes wrong\n\n- **A failed video returns its credits.**\n- **`retry_video`** makes a video again. That is a new charge, so ask first, but it reuses the Look and media so nothing already made is paid for twice.\n- **`cancel_video`** stops a queued or processing video and releases its credits.\n- **Not enough credits**: tell the user their balance and the quote. Plans and credit packs are at https://kinetune.com/pricing. Don't shrink the request without asking.\n- **CLI exit codes**:\n  - `0`: done\n  - `1`: API or network error\n  - `2`: invalid usage, or the quote is over `--max-credits`\n  - `3`: not signed in or not allowed\n- **Unsure of an input?** `kinetune schema \"quote canvas\"` prints the exact JSON schema of any command.\n- **Full reference**: every operation with its parameters, scopes, CLI command and MCP tool is at https://kinetune.com/llms-full.txt.\n"
}