feat: TTS usage v1.5 — upstream-precise (elevenlabs header, gemini usageMetadata)

feat: TTS usage v1 — input character fallback (openai legacy, google, kotodama)

feat: capture Replicate predict_time for image / movie / sound_effect / lipsync agents

feat: surface OpenAI image-gen usage in image_openai_agent

feat(replicate): support image-to-image editing models (Flux Kontext, etc.)

feat: user-configurable concurrency for image/movie/audio generation

feat: add reference image support for imagePrompt in imageParams.images

fix: animated html_tailwind fails when beat.duration is not specified

feat: beat-level local image/movie references (Phase 2)

feat: Swipe風宣言的アニメーション要素をhtml_tailwind beatに追加