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/speech-and-audio.md

View on GitHub
# Speech, Voices, Cloning, and Audio Finishing

Use this guide when working with narration, dialogue, voice cloning, speech conversion, transcription, sound effects, voice isolation, loudness normalization, format conversion, or stem separation.

> Only call MCP tools exposed on this account and listed by the bundled `scripts/mcp_client.py tools` command. Scenario packages include only the subset matched to their workflow; check tool availability before planning a call.

## Discover Capabilities and Live Pricing First

Never hardcode credit numbers, guess durations, or assume every tool uses the same voice features. Query `beatra.models.list` first for live credit rates, supported limits, and capability support.

| User Need | MCP Tool | `beatra.models.list` Filter | Notes |
| --- | --- | --- | --- |
| AI script pacing, pause planning, expressive markup | `beatra.speech.prepare_script` | `{"capability": "speech_script"}` | **Smart Voice Director** — compiles intent into performance-ready script without altering spoken words |
| Single-speaker narration / voiceover | `beatra.speech.synthesize` | `{"capability": "text_to_speech"}` | Optional `direction`, `emotion`, context continuity, free aligned subtitles |
| Native multi-speaker conversation (2–50 turns) | `beatra.speech.synthesize_dialogue` | `{"capability": "text_to_dialogue"}` | Single coherent take with up to 10 `supports_dialogue=true` voices |
| Create a custom cloned voice | `beatra.voices.clone` | `{"capability": "voice_clone"}` | Choose the clone model as §5 describes |
| Activate a previewed voice (paid) | `beatra.voices.activate` | Live price: `activation.price_credits` in the clone result, or the voice in `beatra.voices.list` with `"status": "pending_activation"` | Only for a voice with `status="pending_activation"`; one call charges once (§5 Step E) |
| Speech-to-speech performance conversion | `beatra.speech.convert` | `{"capability": "speech_to_speech"}` | Preserves timing, pacing, and emotion across voices (1 s – 5 min) |
| Audio/video transcription + diarization | `beatra.audio.transcribe` | `{"capability": "speech_to_text"}` | Up to 60 min (`<= 100 MiB`), multi-speaker labels, audio events, subtitles |
| Sound effects and seamless ambient beds | `beatra.audio.generate_sfx` | `{"capability": "sound_effects"}` | 0.5–30 s clips, optional seamless `loop=true` |
| Strip background noise/music from voice | `beatra.audio.process` (`operation="isolate_voice"`) | `{"capability": "audio_voice_isolation"}` | Paid pre-clean for noisy clone samples or voice recordings (4.6 s – 5 min) |
| Split music/mix into vocals and stems | `beatra.audio.process` (`operation="separate_stems"`) | `{"capability": "audio_stem_separation"}` | Match returned `pricing` option by `dimensions.stem_mode` (`"two_stem"` / `"multi_stem"`), or the option without `dimensions` |
| Trim, loop, fade, normalize loudness, convert format | `beatra.audio.process` (`trim`, `loop`, `fade`, `normalize_loudness`, `convert_format`) | `{"capability": "audio_process"}` | Paid audio finishing (includes `ulaw_8k_wav` telephony export) |

### Free / Non-Billable Tools

These tools cost **zero credits** and never charge the wallet:

| MCP Tool | Purpose |
| --- | --- |
| `beatra.speech.validate` | Free preflight for `beatra.speech.synthesize`: returns `valid`, `weighted_characters`, `estimated_duration_seconds`, live `price_credits`, `execution`, and warnings |
| `beatra.speech.validate_dialogue` | Free preflight for `beatra.speech.synthesize_dialogue`: checks turn bounds, character budget, unique/cloned voice limits, `estimated_credits`, `estimated_duration_seconds`, and warnings |
| `beatra.voices.list` | Browse preset voices (866 curated presets) and owned clones with bilingual `search`, facet filters, and `capabilities.*` metadata |
| `beatra.voices.preview` | Return a free audition sample URL (`preview_url`, `kind`) for a ready voice |
| `beatra.voices.update` | Rename or update metadata (`display_name`, `make_default`, `gender`, `age_group`, `use_case`, `persona`, `language`) |
| `beatra.voices.delete` | Two-step permanent voice deletion (`confirmation_token` issued in step 1, valid 15 min) |

Never describe paid finishing or preprocessing steps (`isolate_voice`, `normalize_loudness`, `trim`, `fade`, `loop`, `convert_format`, `separate_stems`, or `prepare_script`) as free.

---

## Production Card, Wallet Checks, and `client_request_id`

Before starting any paid speech, dialogue, cloning, conversion, transcription, SFX, or finishing task:

1. Run any applicable free preflight (`beatra.voices.list` / `beatra.voices.preview` / `beatra.speech.validate` / `beatra.speech.validate_dialogue`) and read live rates via `beatra.models.list`. Keep `beatra.wallet.get` and `beatra.wallet.ledger` as optional read-only tools for answering balance or charge questions — never make `beatra.wallet.get` a mandatory gate before submit, and never ask the user to confirm they have enough credits.
2. Present a concise confirmation summarizing the **six required fields** before the first paid step (after all inputs and parameters are settled):
   - **Work**: exact capability, voice choice, language/duration/format, and any planned finishing step.
   - **Credits**: quoted credits from `beatra.models.list` / preflight (never guess or hardcode prices).
   - **Count**: number of calls/takes/clips planned (a retry with changed parameters counts as a new call).
   - **Identity**: fresh caller-generated `client_request_id` per paid call (minted only once the payload is final).
   - **If stop**: name what the customer already holds and keeps at zero additional charge if they stop here (for example, prepared scripts, auditioned voice picks, preflight duration/credit quotes, or any already-delivered audio; if this is the first paid step, state plainly that nothing has been charged yet). Cancel an already-running job via `beatra.tasks.cancel` only if requested.
   - **If balance short**: relay Beatra's official message and verbatim top-up URL (translated cleanly without altering the URL) and wait until the customer confirms they have topped up before retrying with the same frozen `client_request_id`.
3. Keep one `client_request_id` per paid call. Reuse the **exact same `client_request_id`** only when retrying the **identical payload** after a timeout or unknown transport outcome; generate a **new `client_request_id`** whenever any parameter changes.

---

## 1. Smart Voice Director (`beatra.speech.prepare_script`) & Free Preflight

Replaces legacy manual `/` counting or stripping and guesswork about pauses or expressive markup.

### Workflow

1. Take the customer's plain spoken copy and delivery goals (`instructions`: tone, pacing, dramatic beats, pauses, emotion shifts, pronunciation notes).
2. Call **`beatra.speech.prepare_script`** (`{"capability": "speech_script"}`):
   - Pass `input` (1–50,000 chars), optional `instructions` (<= 2,000 chars), target `voice` and `model` (so formatting adapts to that voice's capabilities), `language` (`"auto"` or explicit locale), optional `context` (`{"before": ..., "after": ...}`), and `client_request_id`.
   - **Word fidelity guarantee**: Smart Voice Director prepares pacing, pauses, and expressive annotations **without altering the user's spoken wording**.
   - **No manual brace-tag authoring**: never manually invent or hand-write braced performance tags in prompts or documentation; rely on `beatra.speech.prepare_script` (and `direction` / `emotion` parameters where supported).
3. Inspect the returned `directed` flag and `scripts[]` (`synthesis`, `summary`, `voice_selection`, `alternatives`, `estimate`, `warnings`):
   - If `synthesis` is `null`, read `warnings` — that chunk cannot be synthesized as-is until adjusted or split.
   - Pass each ready script's **`synthesis` object unchanged** together with your own fresh **`client_request_id`** when calling `beatra.speech.synthesize`.
4. **Always run free `beatra.speech.validate`** with the planned `voice`, prepared `input`, `model`, and delivery parameters before calling `beatra.speech.synthesize`:
   - Verify `valid == true`, review `warnings` (such as pause placement or length advisories), `weighted_characters`, `estimated_duration_seconds`, live `price_credits`, and `execution` (`sync` vs `async`).
   - Adjust breaks or segment long scripts cleanly without manual slash-stripping workarounds.

### Reuse & Concurrency Rules

- **48-hour same-script reuse**: to try another voice within 48 hours for the same wording, instructions, language, and context, **call `beatra.speech.prepare_script` again with the identical arguments changing only `voice` (and `model` if needed)** so Beatra re-compiles markup for the new voice's capabilities using the earlier reading at zero charge. Note: the account balance must still cover the initial text estimate plus any other Smart Voice Director tasks currently in progress.
- **Concurrency cap**: run at most **2 concurrent** `beatra.speech.prepare_script` requests at a time.

---

## 2. Single-Speaker Synthesis (`beatra.speech.synthesize`) & Free Aligned Subtitles

- Pass `voice`, prepared `input`, `model` (`"auto"` unless a capability requires a specific choice), optional `language`, `speed` (`0.5–2.0`), `volume` (`>0–10`), `pitch` (`-1.0–1.0`), `emotion`, `direction`, `previous_text`, `next_text`, `format` (`mp3`/`wav`/`opus`/`pcm`), and `client_request_id`.
- **Keep `subtitle_enable` enabled** (`true` by default; do **not** send `subtitles` on `beatra.speech.synthesize`): `beatra.speech.synthesize` returns **`output.subtitles`** (JSON word/segment timestamps, SRT, and VTT) **free of charge**, plus:
  - `output.audio.speech_start_seconds`
  - `output.audio.speech_end_seconds`
  - `output.audio.integrated_loudness_lufs`
- Use `speech_start_seconds`, `speech_end_seconds`, and `output.subtitles` directly for lip-sync, slide alignment, or assembly (such as `beatra.videos.compose` when available) without paying for a separate transcription pass just to obtain timestamps.

---

## 3. Native Multi-Speaker Dialogue (`beatra.speech.synthesize_dialogue`)

Use **`beatra.speech.synthesize_dialogue`** (`{"capability": "text_to_dialogue"}`) when 2 or more speakers interact in one continuous scene with natural turn-taking, interruptions, laughter, and shared emotional arc.

### Hard Limits & Free Preflight

Always call free **`beatra.speech.validate_dialogue`** before calling **`beatra.speech.synthesize_dialogue`**:

| Constraint | Limit | Action When Exceeded |
| --- | --- | --- |
| Turns per call | **2–50 turns** (`turns[]` with `voice`, `input`, optional `speaker`, `emotion`) | Split into sequential scenes |
| Total characters per call | **<= 2,000 characters** (`compiled_characters` / `characters` in `validate_dialogue`) | Chunk scene-by-scene and pass `previous_text` / `next_text` across boundaries |
| Unique voices per call | **<= 10 voices**, all with `capabilities.supports_dialogue == true` | Fall back to segmented `beatra.speech.synthesize` if cast exceeds 10 |
| Ready cloned voices per call | **<= 4 ready cloned voices** (`category="cloned"`, `status="ready"`) within the 10-voice total | Mix preset dialogue voices or fall back to segmented `beatra.speech.synthesize` |
| Unsupported voice / regional dialect | Any voice with `supports_dialogue == false`, or specialty regional dialect not supported in dialogue | Use segmented `beatra.speech.synthesize` per turn (`previous_text` / `next_text`) and join cleanly in finishing |

- Like single-speaker synthesis, `beatra.speech.synthesize_dialogue` returns **free speaker-attributed `output.subtitles`** (JSON/SRT/VTT) and `segments[]` (`turn_index`, `voice_id`, `speaker`, `text`, `start_seconds`, `end_seconds`) when `subtitle_enable` is `true` (default).

---

## 4. Voice Discovery, Audition, Profile Update & Deletion

### Search & Filter 866 Presets + Owned Clones (`beatra.voices.list`)

- Browse up to 500 voices per page (`limit` <= 500, `cursor` for pagination) across `category="preset"` (866 presets) and `category="cloned"` (`status`: `"ready"`, `"pending_activation"`, `"expired"`, `"any"`). The default is `"ready"`; a voice the user previewed but has not activated is listed only with `"pending_activation"` or `"any"`. With `"any"`, use a `"ready"` voice as it is, activate a `"pending_activation"` one per §5 Step E, and skip an `"expired"` one.
- **Bilingual `search`**: accepts English or Chinese keywords and matches voice names, descriptions, styles, accents, and tags.
- **Query filter flags**: filter directly by `language`, `gender`, `age_group`, `use_case`, `emotion`, `model`, plus capability filters:
  - `supports_dialogue=true`
  - `supports_speech_to_speech=true`
  - `supports_direction=true`
  - `supports_emotion=true`
  - `supports_performance_tags=true`
  - `supports_context=true`
- **All 11 `capabilities.*` metadata fields returned per voice**: inspect before committing a voice to a workflow:
  1. `supports_direction` — honors natural-language `direction` cues in synthesis
  2. `supported_emotions` — supported discrete `emotion` presets
  3. `supports_performance_tags` — voice renders expressive cues compiled by `beatra.speech.prepare_script`
  4. `supported_dialects` — supported regional dialect codes
  5. `supports_context` — supports multi-segment prosody conditioning
  6. `supports_previous_text` — honors top-level `previous_text`
  7. `supports_next_text` — honors top-level `next_text`
  8. `supports_pronunciation_pinyin` — supports pinyin pronunciation compilation via script prep
  9. `supports_pronunciation_ipa` — supports IPA pronunciation compilation via script prep
  10. `supports_speech_to_speech` — eligible as target voice for `beatra.speech.convert`
  11. `supports_dialogue` — eligible for `beatra.speech.synthesize_dialogue`

### Free Audition & Profile Management

- **Audition (`beatra.voices.preview`)**: fetch `preview_url` (`kind`) at zero cost so the user can listen before paying for synthesis or dialogue.
- **Profile & Default Management (`beatra.voices.update`)**: rename (`display_name`), change metadata (`gender`, `age_group`, `use_case` up to 5 tags, `persona` <= 200 chars, `language`), or toggle `make_default=true`/`false` to pin a signature brand/narrator voice.

### Mandatory Confirmation #3 — Two-Step Voice Deletion (`beatra.voices.delete`)

Deleting a cloned voice is permanent (`deleted: true` cannot be undone; using that voice again means cloning it again). Never auto-chain step 1 and step 2 in the same turn without explicit user approval:

1. Call `beatra.voices.delete` with `voice_id` only (**omit `confirmation_token`**).
2. Read the returned `voice`, `paid_slot_warning`, `irreversible_notice`, `confirmation_token`, and `token_expires_at` (**15-minute expiry**). Show these warnings clearly to the user.
3. Only after the user explicitly confirms deletion, call `beatra.voices.delete` with both `voice_id` and `confirmation_token`. If 15 minutes pass before confirmation, repeat step 1 for a fresh token.

---

## 5. Voice Cloning: Sample Cleanup, Model Choice & Mandatory Confirmations

### Step A — Clean Noisy Samples First (When Needed)

If the source recording has background music, street noise, room echo, or overlapping sounds, offer to pre-clean it first with **`beatra.audio.process`** using:
- `operation="isolate_voice"`
- `source`: sample audio ref (`4.6 s – 5 min`, `<= 100 MB`)
- `client_request_id`

Inspect `{"capability": "audio_voice_isolation"}` in `beatra.models.list` for live credits and confirm this **paid** pre-clean step with the user before running it.

### Step B — Mandatory Confirmation #1 (`consent_attested=true`)

Before calling `beatra.voices.clone`, obtain **explicit confirmation from the user** that the recording is their own voice or that they hold verifiable authorization from the voice owner to clone and use it. Never pass `consent_attested=true` without user consent in the current conversation.

### Step C — Selecting a Clone Model

Call `beatra.models.list` with `{"capability": "voice_clone"}`. Each card describes a clone model currently offered; its `pricing`, `constraints` and `constraints.limitations` are authoritative. Answer questions about price, samples or limits from the card, never from memory.

Omit `model` (`auto`) by default; the live clone price is then the price on the card that `auto.default_for_omitted_request` names. Select a specific card only when the job needs multi-speaker dialogue, speech-to-speech conversion or expressive delivery and that card states it and lists the job's language; if no card does, tell the user before cloning. Describe the choice by what it does for the job. On the production card, describe the clone the same way and show the live clone price; add no model ID, provider name, balance figure or free-preview offer.

### Step D — Free Preview Before Purchase

When a card sets `constraints.preview_first_supported` to `true`, the clone can be created as a free preview with `preview_first: true`. Offer it only when the user asks to hear the voice before paying, or when a paid clone is refused and the refusal offers it; never on the production card. Answer questions about preview limits or retention from the card's `limitations`.

### Step E — Mandatory Confirmation #2 (`beatra.voices.activate`)

When a clone result or a listed voice has `status="pending_activation"`:

1. Play the preview to the user and state the date it is kept until (`expires_at`). When the result names a shortfall with a top-up link, relay that sentence with the link unchanged.
2. Show the production card for the activation with the live price: `activation.price_credits` in the clone result, or the voice's entry in `beatra.voices.list` with `{"category": "cloned", "status": "pending_activation"}`. Without a live price, show no figure and read the list first.
3. After the user approves the card, call `beatra.voices.activate` once with `voice_id` and a new `client_request_id`. The call charges once and starts the activation; follow it with `beatra.tasks.get`. A voice that is already ready is not charged.
4. On `insufficient_balance` nothing is charged: relay the message with its top-up link ([Billing, errors, and recovery](billing-errors-and-recovery.md)); after the user says they have topped up, call again with the same `client_request_id`, which never charges twice.
5. If activation fails, the charge is refunded automatically; after it succeeds it is not refunded, and an activated voice that is deleted cannot be restored.

---

## 6. Speech-to-Speech, Transcription, SFX & Audio Finishing

### Speech-to-Speech Performance Conversion (`beatra.speech.convert`)

- `{"capability": "speech_to_speech"}`
- Transfers the speaker's exact timing, pauses, laughter, cadence, and emotional contour from a source recording (`audio` / `source`, **1 s – 5 min**, `<= 50 MB`) into a target voice (`voice` where **`capabilities.supports_speech_to_speech == true`**).
- Key options: `remove_background_noise=true` when the guide vocal is noisy, `format` (`mp3`/`wav`/`opus`/`pcm`), `stability` (`0–1`), `similarity_boost` (`0–1`), `style` (`0–1`), `use_speaker_boost`, `seed`.

### Speech-to-Text Transcription (`beatra.audio.transcribe`)

- `{"capability": "speech_to_text"}`
- Transcribes audio or video up to **60 minutes** (`0.5 s – 3600 s`, `<= 100 MiB`).
- Supports `language` (`"auto"` or explicit locale), **`diarize=true`** with optional `num_speakers` (`1–32`), **`tag_audio_events=true`** (tags laughter, applause, music, sighs, etc.), and **`subtitles=true`** (default) delivering aligned **JSON / SRT / VTT** files with zero extra subtitle charge.

### Sound Effects & Seamless Ambience (`beatra.audio.generate_sfx`)

- `{"capability": "sound_effects"}`
- Generates foley, stingers, UI/transition sounds, or environmental ambience from descriptive `prompt` (`1–2,000` chars) for **`duration_seconds` 0.5 – 30.0 s**.
- Set **`loop=true`** for seamless looping beds (café ambience, rain, room tone, office hum) suitable for background mixing or repeating cleanly via `beatra.audio.process`. Adjust `prompt_influence` (`0.0–1.0`, default `0.3`) and `format` (`mp3`/`wav`/`opus`/`pcm`).

### Audio Finishing & Processing (`beatra.audio.process`)

Always use **`operation`** (not `action`) and **`source`**:

| Operation | Model Lookup (`beatra.models.list`) | Exact Parameters | Notes |
| --- | --- | --- | --- |
| `isolate_voice` | `{"capability": "audio_voice_isolation"}` | `source`, optional `format`, `sample_rate`, `bitrate_kbps`, `channels` | Clean speech from 4.6 s – 5 min (`<= 100 MB`); default output `wav` |
| `separate_stems` | `{"capability": "audio_stem_separation"}` | `source`, `mode` (`"two_stem"` or `"multi_stem"`) only | Read `pricing` option whose `dimensions.stem_mode` matches `mode` (or option without `dimensions`); no `format`/fade args accepted |
| `normalize_loudness` | `{"capability": "audio_process"}` | `source`, `loudness_lufs` (`-30` to `-9`, default `-14`), `true_peak_db` (`-9` to `0`, default `-1`), optional finishing/format fields | Broadcast / platform loudness calibration; default **-14 LUFS** |
| `trim` | `{"capability": "audio_process"}` | `source`, `start_ms`, `end_ms` OR `target_duration_ms` (mutually exclusive) + optional finishing/format fields | Supports fade/loudness/format in same call |
| `loop` | `{"capability": "audio_process"}` | `source`, `target_duration_ms` OR `loop_count` (`2–100`), `loop_start_ms`, `loop_end_ms`, `crossfade_ms` (`50–3000`), `end_with_fade` + optional finishing/format fields | Great for turning a 15–30 s `generate_sfx(loop=true)` bed into a multi-minute bed |
| `fade` | `{"capability": "audio_process"}` | `source`, `fade_in_ms` and/or `fade_out_ms` (`0–30000`), `fade_curve` (`tri`/`exp`/`log`/`qsin`) + optional finishing/format fields | At least one of `fade_in_ms` / `fade_out_ms` required |
| `convert_format` | `{"capability": "audio_process"}` | `source`, **`format`** required (`mp3`, `wav`, `ogg`, `flac`, `m4a`, or **`ulaw_8k_wav`**), optional `sample_rate`, `bitrate_kbps`, `channels` | For telephony **`ulaw_8k_wav`**, omit `sample_rate`/`channels` or pass exactly `sample_rate=8000`, `channels=1` |

Note on limits for `beatra.audio.process` (run one finishing call at a time):
- `isolate_voice`: `4.6 s – 5 min` (`<= 100 MB`).
- `separate_stems`: up to `10 min`, `<= 100 MB`, `20 calls/day`.
- `trim`, `loop`, `fade`, `normalize_loudness`, `convert_format`: sources and outputs up to `60 min` (`<= 300 MB`, minimum result `>= 1 s`), `50 calls/day` (failed attempts count toward the daily cap).

---

## 7. Recoverable Errors (`code` / `details.reason`)

Explain errors to the customer in plain words in their language without exposing tool names, parameter field names, or raw `code` / `details.reason` strings. Keep the full structured error for recovery decisions:

| `code` / `details.reason` | Required Next Action |
| --- | --- |
| `insufficient_balance` | Relay the message with its top-up URL unchanged. If it names another way forward, present that too and follow Billing, errors, and recovery; retry the same request with the same frozen `client_request_id` only after the user says they have topped up and did not choose the other way forward |
| `voice_model_incompatible` / `voice_not_found` / dialogue limits exceeded | Run `beatra.speech.validate_dialogue`; choose `supports_dialogue=true` voices (`<= 10`, `<= 4` clones), chunk `<= 2,000` chars scene-by-scene, or fall back to segmented `beatra.speech.synthesize` |
| `voice_unsupported_for_speech_to_speech` | Pick a voice with `capabilities.supports_speech_to_speech == true` from `beatra.voices.list` |
| `voice_pending_activation` | The voice is a free preview: activate it per §5 Step E (live price on the card, then one `beatra.voices.activate` call after the user approves) before using it for speech |
| `voice_activation_in_progress` / `voice_activation_busy` / `voice_preview_changed` | Nothing more was charged: follow the message (follow the running activation with `beatra.tasks.get`, or call `beatra.voices.activate` again) |
| `invalid_request` (`param="confirmation_token"` on delete) | Restart the deletion with `beatra.voices.delete` step 1 for a new 15-minute token and confirm with the user again |
| Noisy clone sample / clone quality rejection | Propose paid `beatra.audio.process` `operation="isolate_voice"` pre-clean (or request a cleaner dry recording) before re-cloning |
| `idempotency_conflict` / `request_in_progress` / network timeout | Reuse the same `client_request_id` ONLY for the identical request payload (`request_in_progress`: poll existing task instead); generate a new ID if any parameter changed |
| Async job taking long (`in_progress`) / `daily_limit` | Check `beatra.tasks.get` (or `beatra.tasks.list`) and never double-submit; on `daily_limit`, wait until after `00:00 UTC` (`08:00` Beijing time) |
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: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated 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: 194kUpdated Claude CodeCodexVerifiedInstall