keisei/KeiSeiKit-1.0
2
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
KeiSeiKit-1.0/_blocks/api-elevenlabs.md
denis 0b901cf2f9 feat: KeiSeiKit v0.1.0 — initial public release 
Generic Constructor-Pattern agent kit for Claude Code. Zero personal data,
fully English, MIT-licensed.

Contents:
- 34 reusable blocks (baseline, rules, stack/deploy/domain/api/scraper)
- 14 cross-project agent manifests (code/ml/infra/researcher/critic/...)
- 6 portable skills (/new-agent, /research, /test-gen, /debug-deep, /pr-review, /refactor)
- Rust assembler (single binary, ~500 KB)
- 3 hooks (auto-reassemble, pre-commit validate, no-hand-edit)
- install.sh (idempotent, cargo-builds on first run)
- MIT LICENSE

All 6 sanity greps pass: 0 Russian text, 0 specific project names,
0 incident numbers, 0 user paths, 0 hardcoded IPs, 0 API keys.

cargo check + assemble --validate: both pass on 14 manifests.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-20 23:58:34 +08:00

2.5 KiB
Raw Blame History

API — ElevenLabs (voice)

Live pricing: WebFetch https://elevenlabs.io/pricing before any bulk run [VERIFY: character pricing tier varies by plan].

MANDATORY 3-step Voice Design flow (order is fixed):

  1. designVoice — describe voice characteristics (gender, age, accent, style) → returns preview audio + generated_voice_id (ephemeral).
  2. createVoice — accept the preview → permanent voice_id added to library.
  3. TTS — synthesize text using the permanent voice_id.

Skipping or reordering any step = API error. Ephemeral preview IDs expire — cannot TTS directly from designVoice output.

Models:

Model Use case Latency Quality
eleven_flash_v2_5 Real-time, low latency (~75ms) Fastest Good
eleven_multilingual_v2 Production, 29 languages Slower Best
eleven_turbo_v2_5 Balanced Fast High

Pricing [VERIFY: check live pricing page] — billed per character, plan-gated character quota:

  • Free: ~10K chars/mo
  • Starter: ~30K chars/mo
  • Creator / Pro / Scale — higher quotas, character overage rates vary per plan.
  • Voice Design calls also consume characters (preview audio counts).

TTS params (sane defaults):

  • stability: 0.5 — higher = more monotone, lower = more expressive (range 0-1)
  • similarity_boost: 0.75 — higher = closer to reference voice
  • style: 0-1 — emotional exaggeration; set 0 for Flash v2 (not supported)
  • use_speaker_boost: true for Multilingual v2

Voice ID caching: once createVoice returns a voice_id, store it in memory/{project}.md or DB. Reuse across TTS calls — re-designing the same voice = wasted characters + non-deterministic result.

Video integration (if pairing with a video model that supports voice): voice_id flows into the video model's voice_ids payload. Per-speaker markers in prompts ONLY when voice_ids actually sent.

Cost tracking: log per-call characters_used + cumulative month-to-date → memory/{project}.md. Hand off to cost-guardian on any batch expected to exceed 50% of monthly quota.

Forbidden: calling TTS without prior createVoice (ephemeral preview IDs fail); exceeding plan character quota without cost-guardian check (overage billing surprise); committing voice_id values into git when they reference private/cloned voices (storage convention — see domain-has-secrets.md); re-designing the same voice per-scene instead of caching voice_id; skipping the 3-step flow with direct TTS on generated_voice_id.