Skip to content

beatra-ai/beatra-skills / plugins/beatra/skills/beatra

beatra

Verified

Create AI images, videos, music, and voice, look up public social data, edit visual results, and keep generated assets easy to find and manage. Beatra Universal turns text into images, images and references into video, ideas into songs or instrumentals, and scripts into narration, while also supporting custom voice creation and public social media data lookup across YouTube, TikTok, Instagram, and X. Use it as an AI media generator and AI creative suite for text-to-image, image-to-video, AI video editing, AI music generation, text-to-speech, multilingual voice-over, voice cloning, multimodal content creation, and YouTube or TikTok data for posts, comments, accounts, and trends. Reuse source media, returned files, and public social findings across formats, follow production progress, and view actual credit use for completed creative work.

references/music.md

View on GitHub
# Music

Use `beatra.music.generate` for a song with lyrics, an instrumental, or a new
track guided by reference audio.

- For an instrumental, put genre, mood, tempo feel, instrumentation, structure,
  and intended use in the prompt. Keep lyrics absent.
- When the user supplies lyrics, preserve them exactly unless they ask for
  writing or revision. Use the prompt for musical and vocal direction.
- For reference-guided work, state what musical qualities should guide the new
  result. Do not promise exact melody, voice, identity, or arrangement
  preservation.
- Name musical qualities, not recording artists. When the music supplier flags a
  prompt, tag, or lyric for naming a real artist, it returns a terminal
  `safety_rejected` task; the prepaid credits are refunded, but the attempt is
  wasted. When the user names an artist as a reference, keep their intent and
  describe what they are reaching for — genre, era, tempo feel, instrumentation,
  vocal texture, production style — rather than passing the name through. Tell
  the user once that the reference was expressed as musical qualities, so they
  know what was actually sent.

Omit `model` unless the user explicitly chooses one. Use
`beatra.models.list` only when current compatibility, supported controls,
constraints, or price matters:

```bash
python3 scripts/mcp_client.py call beatra.models.list
```

```json
{"capability": "text_to_music"}
```

For a reference-guided track use `{"capability":"reference_audio_to_music"}`.
Model-specific options are accepted only when the returned interface card
documents them; never move an option between model families or silently drop
one.

For a local reference, use only:

```text
python3 scripts/mcp_client.py upload <path> --mime-type <type>
```

Use its returned artifact as `reference_audio`. Do not manually call the raw
upload grant, use host HTTP, or send a local path. Respect the general 100 MB
upload ceiling plus any lower current model-specific limit.

The bundled MCP music route does not accept REST-only `callback_url`,
`callback_signing_key_id`, or `metadata`. Do not add them or use another
transport to obtain them.

A sung track needs `prompt`, `lyrics`, `title`, and one stable
`client_request_id`. `title` stops being optional the moment lyrics are
supplied.

```bash
python3 scripts/mcp_client.py call beatra.music.generate
```

```json
{
  "prompt": "Warm indie folk, fingerpicked guitar, unhurried, for a family photo montage",
  "lyrics": "Morning finds the window first\nand waits there while we wake",
  "title": "Morning Finds the Window",
  "client_request_id": "music-sung-opaque-1"
}
```

Music for a video or a fixed slot is requested at that length. Without
`duration_seconds` a track runs about three minutes, and a join under a
15-second clip keeps only its opening. Send `duration_seconds` near the video's
length: an instrumental bed for a 15-second ad takes `15` with
`duration_fit: "exact"`; a sung track takes 30 or more that its lyrics fit,
so under a video shorter than 30 seconds only its opening plays: say so, or
suggest an instrumental. The
join already fades music out at the video's end, so a slightly longer bed is
fine. The rules and limits are in the length section of
[music editing and audio](music-editing-and-audio.md).

An instrumental sets `instrumental` and sends no lyrics; a title stays welcome
when the user named one. Lyrics and `instrumental: true` exclude each other, and
a request carrying both is rejected.

```json
{
  "prompt": "Warm indie folk, fingerpicked guitar, unhurried, for a family photo montage",
  "instrumental": true,
  "client_request_id": "music-instrumental-opaque-1"
}
```

Omitting `model` routes the request portably. The portable route holds the
prompt and lyrics to its own bounds, which are published on no card and are not
the bounds of whichever model ends up serving the request; a card's limits
govern only the model it belongs to, and only when that model is named. So when
a prompt or lyric is long enough that a limit could matter, name the model whose
card admits that length instead of sending it portably. A portable request
refused for length names the exact bound it applied in the returned message.

Music generation is one billable asynchronous request. Finalize the complete
prompt, lyrics, instrumental status, title, model, reference, and accepted model
options before creating the stable `client_request_id`. Submit once and poll
the same task. An identical retry preserves every validated generation
argument; any accepted change is new paid work with a new identity and
confirmation.

On success, deliver every returned clip in order; one generation usually
returns two versions of the same song, so say that the user can pick one. Include
returned title and lyrics when present plus each audio link, duration, format,
and size when returned, in plain words. Report the credits charged in plain
words; keep artifact IDs, MIME strings, and the resolved model in your notes
unless the user asks. Do not claim to have reviewed composition, vocals, pronunciation, or
mix when the host cannot play the audio.

## Singers

Four tools carry one singer across songs. They exist only when the bundled
client's `tools` command lists them.

- `beatra.music.personas.list` lists the singers saved on this account with
  their `persona_id` values.
- `beatra.music.personas.create_from_song` saves the singer of a song Beatra
  generated as a reusable singer (paid, on its own card).
- `beatra.music.personas.enroll_voice` takes the user's own a cappella recording
  and returns the verification phrase the user records (free).
- `beatra.music.personas.verify_voice` checks the user's recording of that
  phrase and activates a singer in the user's own voice (paid, on its own card).

When the user asks for a song in their own voice, use the singer already made
from their voice (see "Saved singers" in
[music editing and audio](music-editing-and-audio.md)); if there is none,
create one first ("A singer from the user's own voice"). Those sections also
cover how a singer's `persona_id` reaches `beatra.music.generate`.
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 196kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 196kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 196kUpdated Claude CodeCodexVerifiedInstall