videobook 0.0.0-stage → 0.2.0

package/dist/media/adapters/xai.js

@@ -0,0 +1,18 @@
+import { requireApiKey } from "../../api-key.js";
+import { uploadAssetFile } from "../../upload.js";
+import { callXaiPaid, createXaiPaygatedProvider, XAI_VIDEO_GENERATIONS_DOT_PATH } from "../../jobs/xai-paygate.js";
+import { writePendingTask } from "./pending.js";
+import { MEDIA_TASK_TYPES } from "./types.js";
+async function submit(request, context) {
+    const provider = createXaiPaygatedProvider(requireApiKey("XAI_API_KEY"), 120_000);
+    const { request_id } = await callXaiPaid(provider.post.v1.videos.generations, XAI_VIDEO_GENERATIONS_DOT_PATH, request);
+    await writePendingTask(context, request_id, MEDIA_TASK_TYPES.xai);
+    return { kind: "pending", taskId: request_id };
+}
+export const xaiAdapter = {
+    provider: "xai",
+    envKey: "XAI_API_KEY",
+    upload: uploadAssetFile,
+    uploadUrl: async (url) => url,
+    submit,
+};

package/dist/media/catalog.js

@@ -0,0 +1,236 @@
+import { MiniMaxH3ReferenceClipDurationMs } from "@apicity/kie/zod";
+import { LIGHTNING_SCENE_OPTIONS, LIGHTNING_VIDEO_OPTIONS } from "../lightning-defaults.js";
+const SEEDANCE_MODES = [
+    { id: "t2v", label: "Text to video", roles: [] },
+    { id: "i2v", label: "Frames", roles: ["first_frame", "last_frame"], requireAny: ["first_frame", "last_frame"] },
+    { id: "r2v", label: "References", roles: ["image", "video", "audio"], requireAny: ["image", "video", "audio"] },
+];
+const FRAME_MODES = [
+    { id: "i2v", label: "Frames", roles: ["first_frame", "last_frame"] },
+];
+const SEEDANCE_US_REFERENCE_MODES = [
+    { id: "r2v", label: "References", roles: ["image", "video", "audio"], requireAny: ["image", "video", "audio"] },
+    { id: "v2v", label: "Edit video", roles: ["video", "image", "audio"], requireAny: ["video"], defaults: { task: "editing" } },
+    { id: "extend", label: "Extend video", roles: ["video", "image", "audio"], requireAny: ["video"], defaults: { task: "extension" } },
+];
+const XAI_MODES = [
+    { id: "i2v", label: "Image to video", roles: ["first_frame"] },
+    { id: "r2v", label: "References", roles: ["image"] },
+    { id: "v2v", label: "Edit video", roles: ["video"], requireAny: ["video"] },
+];
+const SPEAKER = "Speaker 1";
+function kieGeminiClassicTts(id, title) {
+    return {
+        id,
+        media: "audio",
+        operations: ["speech"],
+        title,
+        prompt: "dialogue_turns.0.text",
+        aliases: {
+            voice: "speakers.0.voice_name",
+            accent: "speakers.0.accent",
+            style: "speakers.0.style",
+            pace: "speakers.0.pace",
+        },
+        defaults: {
+            speakers: [{ speaker_id: SPEAKER, voice_name: "Kore", accent: "Neutral" }],
+            dialogue_turns: [{ speaker_id: SPEAKER }],
+        },
+    };
+}
+function kieGemini38Tts(id, title) {
+    return {
+        id,
+        media: "audio",
+        operations: ["speech"],
+        title,
+        prompt: "dialogue_turns.0.text",
+        aliases: { voice: "speakers.0.voice_name", style: "dialogue_turns.0.style" },
+        defaults: {
+            speakers: [{ speaker_id: SPEAKER, voice_name: "Fola" }],
+            dialogue_turns: [{ speaker_id: SPEAKER }],
+        },
+    };
+}
+function kieElevenLabsTts(id, title) {
+    return { id, media: "audio", operations: ["speech"], title, defaults: { voice: "Rachel" } };
+}
+function h3MaxStyle(style) {
+    return { id: `fal:minimax/h3-max/styles/${style}`, media: "video", operations: ["text_to_video", "image_to_video"], dialect: "strip_frames" };
+}
+export const MODEL_ENTRIES = [
+    { id: "kie:nano-banana-2-1", media: "image", operations: ["text_to_image", "edit_image"], dialect: "natural_numbered", defaults: { ...LIGHTNING_SCENE_OPTIONS } },
+    { id: "kie:nano-banana-pro", media: "image", operations: ["text_to_image", "edit_image"], guide: "nano_banana", dialect: "natural" },
+    { id: "kie:gpt-image-2-5-flare-text-to-image", media: "image", operations: ["text_to_image"], guide: "gpt_image", dialect: "natural", defaults: { resolution: "2K" } },
+    { id: "kie:gpt-image-2-5-flare-image-to-image", media: "image", operations: ["edit_image"], guide: "gpt_image", dialect: "natural", defaults: { resolution: "2K" } },
+    { id: "kie:gpt-image-2-5-sunburst-text-to-image", media: "image", operations: ["text_to_image"], guide: "gpt_image", dialect: "natural", defaults: { resolution: "2K" } },
+    { id: "kie:gpt-image-2-5-sunburst-image-to-image", media: "image", operations: ["edit_image"], guide: "gpt_image", dialect: "natural", defaults: { resolution: "2K" } },
+    { id: "kie:gpt-image-2-image-to-image", media: "image", operations: ["edit_image"], title: "GPT Image 2 Edit", dialect: "natural", hidden: true },
+    { id: "kie:seedream/5-pro-image-to-image", media: "image", operations: ["edit_image"], dialect: "natural", hidden: true },
+    { id: "kie:grok-imagine/text-to-image", media: "image", operations: ["text_to_image"], guide: "grok_image" },
+    { id: "kie:grok-imagine/image-to-image", media: "image", operations: ["edit_image"], guide: "grok_image", dialect: "natural" },
+    { id: "kie:qwen2/image-edit", media: "image", operations: ["edit_image"], dialect: "natural" },
+    { id: "kie:qwen3/image-to-image", media: "image", operations: ["edit_image"], dialect: "natural" },
+    { id: "kie:qwen2-1/text-to-image", media: "image", operations: ["text_to_image"], dialect: "natural" },
+    { id: "kie:qwen2-1/image-to-image", media: "image", operations: ["edit_image"], dialect: "natural" },
+    { id: "kie:seedream/5-lite-text-to-image", media: "image", operations: ["text_to_image"] },
+    { id: "kie:seedream/5-lite-image-to-image", media: "image", operations: ["edit_image"], dialect: "natural" },
+    { id: "kie:seedream/5-flash-text-to-image", media: "image", operations: ["text_to_image"] },
+    { id: "kie:seedream/5-flash-image-to-image", media: "image", operations: ["edit_image"] },
+    { id: "kie:seedream/5-flash-layer-decomposition", media: "image", operations: ["layer_image"] },
+    { id: "alibaba:qwen-image-2.0-pro", media: "image", operations: ["text_to_image"], title: "Qwen Image 2.0 Pro (Alibaba)" },
+    { id: "alibaba:qwen-image-edit-plus", media: "image", operations: ["edit_image"], title: "Qwen Image Edit Plus (Alibaba)" },
+    { id: "fal:fal-ai/nano-banana", media: "image", operations: ["text_to_image"], guide: "nano_banana", hidden: true },
+    { id: "fal:fal-ai/nano-banana/edit", media: "image", operations: ["edit_image"], guide: "nano_banana", hidden: true },
+    { id: "fal:fal-ai/nano-banana-2", media: "image", operations: ["text_to_image"], guide: "nano_banana", hidden: true },
+    { id: "fal:fal-ai/nano-banana-2/edit", media: "image", operations: ["edit_image"], guide: "nano_banana", hidden: true },
+    { id: "fal:google/nano-banana-2-lite", media: "image", operations: ["text_to_image"], guide: "nano_banana", hidden: true },
+    { id: "fal:google/nano-banana-lite/edit", media: "image", operations: ["edit_image"], guide: "nano_banana", hidden: true },
+    { id: "fal:fal-ai/nano-banana-pro", media: "image", operations: ["text_to_image"], guide: "nano_banana", hidden: true },
+    { id: "fal:fal-ai/nano-banana-pro/edit", media: "image", operations: ["edit_image"], guide: "nano_banana", hidden: true },
+    { id: "fal:google/nano-banana-2.1", media: "image", operations: ["text_to_image"], guide: "nano_banana", hidden: true },
+    { id: "fal:google/nano-banana-2.1/edit", media: "image", operations: ["edit_image"], guide: "nano_banana", hidden: true },
+    { id: "fal:fal-ai/bytedance/seedream/v5/lite/text-to-image", media: "image", operations: ["text_to_image"], hidden: true },
+    { id: "fal:fal-ai/bytedance/seedream/v5/lite/edit", media: "image", operations: ["edit_image"], hidden: true },
+    { id: "fal:xai/grok-imagine-image", media: "image", operations: ["text_to_image"], guide: "grok_image", hidden: true },
+    { id: "fal:xai/grok-imagine-image/edit", media: "image", operations: ["edit_image"], guide: "grok_image", hidden: true },
+    { id: "fal:fal-ai/qwen-image", media: "image", operations: ["text_to_image"], hidden: true },
+    { id: "fal:fal-ai/qwen-image-edit", media: "image", operations: ["edit_image"], hidden: true },
+    { id: "fal:alibaba/qwen-image-3/edit", media: "image", operations: ["edit_image"] },
+    { id: "fal:fal-ai/wan/v2.7/edit", media: "image", operations: ["edit_image"], rules: { maxPromptChars: 500 }, hidden: true },
+    { id: "fal:bria/fibo-edit-1.5/product-holding", media: "image", operations: ["edit_image"], prompt: "instruction", refs: { image: ["person_image_url", "product_image_urls"] } },
+    { id: "fal:bria/fibo-edit-1.5/virtual-try-on", media: "image", operations: ["edit_image"], prompt: "instruction", refs: { image: ["person_image_url", "garment_image_urls"] } },
+    { id: "fal:recraft/v4.1/flash/text-to-image", media: "image", operations: ["text_to_image"] },
+    { id: "fal:bytedance/seedream/v5/flash/text-to-image", media: "image", operations: ["text_to_image"] },
+    { id: "fal:bytedance/seedream/v5/flash/edit", media: "image", operations: ["edit_image"] },
+    { id: "fal:bytedance/seedream/v5/flash/layerize", media: "image", operations: ["layer_image"] },
+    { id: "fal:ideogram/v4.5", media: "image", operations: ["text_to_image"] },
+    { id: "fal:ideogram/v4.5/edit", media: "image", operations: ["edit_image"], refs: { image: ["image_url", "reference_image_urls"] } },
+    { id: "fal:blackforestlabs/flux-3/text-to-image", media: "image", operations: ["text_to_image"] },
+    { id: "fal:blackforestlabs/flux-3/edit-image", media: "image", operations: ["edit_image"] },
+    { id: "fal:fal-ai/flux-2-flex", media: "image", operations: ["text_to_image"] },
+    { id: "fal:bria/guided-remove-background", media: "image", operations: ["edit_image"] },
+    { id: "fal:bria/fibo-edit-1.5/expand", media: "image", operations: ["edit_image"] },
+    { id: "fal:fal-ai/marigold-v2", media: "image", operations: ["edit_image"] },
+    {
+        id: "kie:bytedance/seedance-2-5",
+        media: "video",
+        operations: ["text_to_video", "image_to_video", "reference_to_video"],
+        guide: "seedance25",
+        dialect: "seedance",
+        modes: SEEDANCE_MODES,
+        defaults: { ...LIGHTNING_VIDEO_OPTIONS },
+        frameOptions: { aspect_ratio: "adaptive" },
+    },
+    { id: "kie:grok-imagine/image-to-video", media: "video", operations: ["image_to_video"], guide: "grok_imagine" },
+    { id: "kie:grok-imagine/text-to-video", media: "video", operations: ["text_to_video"], guide: "grok_imagine", defaults: { duration: 6, resolution: "480p" } },
+    { id: "kie:grok-imagine/extend", media: "video", operations: ["extend_video"], hidden: true },
+    { id: "kie:grok-imagine/upscale", media: "video", operations: ["upscale_video"], hidden: true },
+    { id: "kie:kling-3.0-omni/text-to-video", media: "video", operations: ["text_to_video"], guide: "kling30", dialect: "natural", defaults: { customize_multi_shots: false } },
+    { id: "kie:kling-3.0-omni/image-to-video", media: "video", operations: ["image_to_video"], guide: "kling30", dialect: "natural" },
+    { id: "kie:kling-3.0-omni/reference-to-video", media: "video", operations: ["reference_to_video"], guide: "kling30", dialect: "natural", defaults: { audio: false } },
+    {
+        id: "kie:kling-3.0-omni/transformation",
+        media: "video",
+        operations: ["edit_video"],
+        guide: "kling30",
+        dialect: "natural",
+        modes: [
+            { id: "v2v", label: "Edit video", roles: ["video"], requireAny: ["video"], defaults: { aspect_ratio: "auto" } },
+            { id: "v2v_refs", label: "Edit with references", roles: ["video", "image"], requireAny: ["video"] },
+        ],
+    },
+    { id: "kie:kling-3.0/motion-control", media: "video", operations: ["motion_control"], guide: "kling30", dialect: "natural", defaults: { mode: "720p" } },
+    { id: "kie:gemini-omni-video", media: "video", operations: ["image_to_video", "edit_video"], dialect: "natural_source_video", defaults: { duration: "8" }, modes: [{ id: "i2v", label: "Image to video", roles: ["image"] }, { id: "v2v", label: "Edit video", roles: ["video", "image"], requireAny: ["video"] }], refItem: { video: { start: 0, ends: 8 } }, aliases: { clip_start: "video_list.0.start", clip_end: "video_list.0.ends" } },
+    { id: "kie:wan/3-0-video", media: "video", operations: ["text_to_video", "image_to_video", "reference_to_video"], dialect: "seedance", modes: SEEDANCE_MODES },
+    { id: "kie:wan/3-0-video-prime", media: "video", operations: ["text_to_video", "image_to_video", "reference_to_video"], dialect: "seedance", modes: SEEDANCE_MODES },
+    { id: "kie:happyhorse-1-1/text-to-video", media: "video", operations: ["text_to_video"] },
+    { id: "kie:happyhorse-1-1/image-to-video", media: "video", operations: ["image_to_video"], dialect: "strip_frames" },
+    { id: "kie:happyhorse-1-1/reference-to-video", media: "video", operations: ["reference_to_video"], dialect: "brackets" },
+    { id: "kie:happyhorse/video-edit", media: "video", operations: ["edit_video"], dialect: "happyhorse_edit" },
+    { id: "kie:minimax-h3/text-to-video", media: "video", operations: ["text_to_video"] },
+    { id: "kie:minimax-h3/image-to-video", media: "video", operations: ["image_to_video"], dialect: "strip_frames" },
+    { id: "kie:minimax-h3/reference-to-video", media: "video", operations: ["reference_to_video"], dialect: "natural", rules: { refDurationMs: MiniMaxH3ReferenceClipDurationMs } },
+    { id: "kie:volcengine/video-to-video-lip-sync", media: "video", operations: ["lip_sync"] },
+    { id: "kie:omnihuman-1-5", media: "video", operations: ["image_to_video", "lip_sync"], dialect: "natural" },
+    { id: "kie:vidu-q4/image-to-video", media: "video", operations: ["image_to_video"], dialect: "natural_kinds" },
+    { id: "kie:vidu-q4/reference-to-video", media: "video", operations: ["reference_to_video"], dialect: "natural_kinds" },
+    { id: "kie:veo3", media: "video", operations: ["text_to_video", "image_to_video"], title: "Veo 3 (Kie)" },
+    { id: "kie:veo3_fast", media: "video", operations: ["text_to_video", "image_to_video"], title: "Veo 3 Fast (Kie)" },
+    { id: "xai:grok-imagine-video", media: "video", operations: ["text_to_video", "image_to_video", "reference_to_video", "edit_video"], guide: "grok_imagine", title: "Grok Imagine Video (xAI)", dialect: "natural", refs: { first_frame: "image", audio: null }, modes: XAI_MODES },
+    { id: "fal:xai/grok-imagine-video/image-to-video", media: "video", operations: ["image_to_video"], guide: "grok_imagine", hidden: true },
+    { id: "fal:xai/grok-imagine-video/reference-to-video", media: "video", operations: ["reference_to_video"], guide: "grok_imagine", hidden: true },
+    { id: "fal:xai/grok-imagine-video/extend-video", media: "video", operations: ["extend_video"], hidden: true },
+    { id: "fal:xai/grok-imagine-video/edit-video", media: "video", operations: ["edit_video"], guide: "grok_imagine", hidden: true },
+    { id: "fal:xai/grok-imagine-video/v1.5/lite/image-to-video", media: "video", operations: ["image_to_video"] },
+    { id: "fal:xai/grok-imagine-video/v1.5/lite/text-to-video", media: "video", operations: ["text_to_video"] },
+    { id: "fal:alibaba/wan-3.0/text-to-video", media: "video", operations: ["text_to_video"], hidden: true },
+    { id: "fal:alibaba/wan-3.0/image-to-video", media: "video", operations: ["image_to_video"], hidden: true },
+    { id: "fal:alibaba/wan-3.0/reference-to-video", media: "video", operations: ["reference_to_video"], hidden: true },
+    { id: "fal:alibaba/wan-3.0-prime/text-to-video", media: "video", operations: ["text_to_video"], hidden: true },
+    { id: "fal:alibaba/wan-3.0-prime/image-to-video", media: "video", operations: ["image_to_video"], hidden: true },
+    { id: "fal:alibaba/wan-3.0-prime/reference-to-video", media: "video", operations: ["reference_to_video"], hidden: true },
+    { id: "fal:minimax/h3-max/insert-video", media: "video", operations: ["edit_video"], refs: { video: ["video_url", "reference_video_urls"] } },
+    { id: "fal:minimax/h3-max/recast", media: "video", operations: ["edit_video"] },
+    { id: "fal:minimax/h3-max-turbo/extend-video", media: "video", operations: ["extend_video"] },
+    { id: "fal:minimax/h3-max/relight", media: "video", operations: ["edit_video"] },
+    h3MaxStyle("16bit-pixel"),
+    h3MaxStyle("hand-drawn"),
+    h3MaxStyle("low-poly"),
+    h3MaxStyle("retro-toon-70s"),
+    h3MaxStyle("vhs"),
+    { id: "fal:bytedance/seedance-2.5/us/text-to-video", media: "video", operations: ["text_to_video"], guide: "seedance25", dialect: "seedance", hidden: true },
+    { id: "fal:bytedance/seedance-2.5/us/image-to-video", media: "video", operations: ["image_to_video"], guide: "seedance25", dialect: "seedance", refs: { first_frame: "image_url" }, modes: FRAME_MODES, hidden: true },
+    { id: "fal:bytedance/seedance-2.5/us/reference-to-video", media: "video", operations: ["reference_to_video", "edit_video", "extend_video"], guide: "seedance25", dialect: "seedance", modes: SEEDANCE_US_REFERENCE_MODES, hidden: true },
+    { id: "fal:bytedance/seedance-2.5/draft/complete", media: "video", operations: ["upscale_video"], title: "Seedance 2.5 Draft Complete", hidden: true },
+    { id: "fal:bytedance/seedance-2.0/us/text-to-video", media: "video", operations: ["text_to_video"], guide: "seedance2", dialect: "seedance" },
+    { id: "fal:bytedance/seedance-2.0/us/image-to-video", media: "video", operations: ["image_to_video"], guide: "seedance2", dialect: "seedance", refs: { first_frame: "image_url" }, modes: FRAME_MODES },
+    { id: "fal:bytedance/seedance-2.0/us/reference-to-video", media: "video", operations: ["reference_to_video"], guide: "seedance2", dialect: "seedance", modes: [{ id: "r2v", label: "References", roles: ["image", "video", "audio"], requireAny: ["image", "video"] }] },
+    { id: "fal:fal-ai/kling-video/o3/standard/image-to-video", media: "video", operations: ["image_to_video"], guide: "kling30", dialect: "strip_frames", refs: { first_frame: "image_url" }, modes: FRAME_MODES },
+    { id: "fal:fal-ai/kandinsky6-pro/text-to-video", media: "video", operations: ["text_to_video"] },
+    { id: "fal:fal-ai/kandinsky6-pro/image-to-video", media: "video", operations: ["image_to_video"], dialect: "strip_frames" },
+    { id: "fal:fal-ai/kandinsky6-lite/text-to-video", media: "video", operations: ["text_to_video"] },
+    { id: "fal:fal-ai/kandinsky6-lite/image-to-video", media: "video", operations: ["image_to_video"], dialect: "strip_frames" },
+    { id: "fal:fal-ai/kandinsky6-vsr", media: "video", operations: ["upscale_video"] },
+    { id: "fal:fal-ai/kandinsky6-vsr/lite", media: "video", operations: ["upscale_video"] },
+    { id: "fal:fal-ai/vidu/q4/image-to-video", media: "video", operations: ["image_to_video"], dialect: "natural_kinds", hidden: true },
+    { id: "fal:fal-ai/vidu/q4/reference-to-video", media: "video", operations: ["reference_to_video"], dialect: "vidu", hidden: true },
+    { id: "fal:pixverse/music-video/vibemv", media: "video", operations: ["reference_to_video"], refs: { image: ["image_url", "style_image_url"] } },
+    { id: "fal:pixelcut/looping-video", media: "video", operations: ["image_to_video"], dialect: "strip_frames" },
+    { id: "fal:creatify/boreal", media: "video", operations: ["text_to_video", "image_to_video", "lip_sync"] },
+    { id: "fal:mirelo-ai/sfx2.0/video-to-video", media: "video", operations: ["edit_video"], prompt: "text_prompt", defaults: { num_samples: 1 } },
+    { id: "elevenlabs:text-to-speech", media: "audio", operations: ["speech"], title: "ElevenLabs Text to Speech", defaults: { model_id: "eleven_multilingual_v2" } },
+    { id: "elevenlabs:sound-generation", media: "audio", operations: ["sfx"], title: "ElevenLabs Sound Effects" },
+    { id: "elevenlabs:music", media: "audio", operations: ["music"], title: "ElevenLabs Music", defaults: { force_instrumental: true } },
+    { id: "elevenlabs:video-to-music", media: "audio", operations: ["video_to_music"], title: "ElevenLabs Video to Music" },
+    { id: "kie:elevenlabs/sound-effect-v2", media: "audio", operations: ["sfx"] },
+    kieElevenLabsTts("kie:elevenlabs/text-to-speech-multilingual-v2", "ElevenLabs Multilingual v2 (Kie)"),
+    kieElevenLabsTts("kie:elevenlabs/text-to-speech-turbo-2-5", "ElevenLabs Turbo 2.5 (Kie)"),
+    {
+        id: "kie:elevenlabs/text-to-dialogue-v3",
+        media: "audio",
+        operations: ["dialogue"],
+        title: "ElevenLabs v3 Dialogue (Kie)",
+        prompt: "dialogue.0.text",
+        aliases: { voice: "dialogue.0.voice" },
+        defaults: { dialogue: [{ voice: "Rachel" }] },
+    },
+    kieGeminiClassicTts("kie:google/gemini-2-5-pro-tts", "Gemini 2.5 Pro TTS (Kie)"),
+    kieGeminiClassicTts("kie:google/gemini-3-1-flash-tts", "Gemini 3.1 Flash TTS (Kie)"),
+    kieGemini38Tts("kie:google/gemini-3-8-flash-tts", "Gemini 3.8 Flash TTS (Kie)"),
+    kieGemini38Tts("kie:google/gemini-3-8-flash-lite-tts", "Gemini 3.8 Flash Lite TTS (Kie)"),
+    { id: "kie:suno/generate", media: "audio", operations: ["music"], title: "Suno (Kie)", defaults: { model: "V5_5", instrumental: true, customMode: false } },
+    { id: "kie:suno/sounds-generate", media: "audio", operations: ["sfx"], title: "Suno Sounds (Kie)", defaults: { model: "V5_5" } },
+    { id: "fal:elevenlabs/music/v2.5", media: "audio", operations: ["music"], defaults: { force_instrumental: true, output_format: "mp3_48000_192" } },
+    { id: "fal:minimax/music-3", media: "audio", operations: ["music"], defaults: { lyrics: "[Instrumental]" } },
+    { id: "fal:elevenlabs/tts/eleven-v4", media: "audio", operations: ["speech"], defaults: { voice: "Rachel", output_format: "mp3_44100_128" } },
+    { id: "fal:elevenlabs/tts/eleven-v4-turbo", media: "audio", operations: ["speech"], defaults: { voice: "Rachel", output_format: "mp3_44100_128" } },
+    { id: "fal:google/gemini-3.8-flash-tts", media: "audio", operations: ["speech"], defaults: { voice: "Kore" } },
+    { id: "fal:google/gemini-3.8-flash-lite-tts", media: "audio", operations: ["speech"], defaults: { voice: "Kore" } },
+    { id: "fal:alibaba/qwen-audio-3-tts", media: "audio", operations: ["speech"], defaults: { voice: "Cherry", language: "Auto" } },
+    { id: "fal:fal-ai/bytedance/seed-speech/tts/v2", media: "audio", operations: ["speech"], defaults: { output_format: "mp3" } },
+];
+const ENTRIES_BY_ID = new Map(MODEL_ENTRIES.map((entry) => [entry.id, entry]));
+export function modelEntry(id) {
+    return ENTRIES_BY_ID.get(id);
+}

package/dist/media/complete.js

@@ -0,0 +1,191 @@
+import { mkdtemp, readFile, rm, writeFile } from "node:fs/promises";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { queue } from "../queue.js";
+import { contentHash } from "../upload.js";
+import { probeVideo, runFfmpeg } from "../ffmpeg.js";
+import { record } from "./adapters/pending.js";
+const EXTENSIONS = {
+    image: ["png", "jpg", "jpeg", "webp"],
+    video: ["mp4", "mov", "webm"],
+    audio: ["mp3", "wav", "ogg", "opus", "flac", "m4a", "aac"],
+};
+const FALLBACK_EXTENSION = { image: "png", video: "mp4", audio: "mp3" };
+function mediaOf(meta) {
+    const media = meta.media;
+    if (media === "image" || media === "video" || media === "audio")
+        return media;
+    throw new Error(`Media generation task is missing meta.media`);
+}
+function extensionFromUrl(url, media) {
+    try {
+        const match = /\.([a-z0-9]{2,5})$/i.exec(new URL(url).pathname);
+        const extension = match?.[1]?.toLowerCase();
+        return extension && EXTENSIONS[media].includes(extension) ? extension : null;
+    }
+    catch {
+        return null;
+    }
+}
+function extensionFromContentType(contentType, media) {
+    const subtype = contentType?.split(";")[0]?.split("/")[1]?.toLowerCase();
+    if (!subtype)
+        return null;
+    const normalized = subtype === "mpeg" ? "mp3" : subtype === "quicktime" ? "mov" : subtype === "x-wav" ? "wav" : subtype;
+    return EXTENSIONS[media].includes(normalized) ? normalized : null;
+}
+async function download(output, media) {
+    if ("bytes" in output)
+        return { bytes: output.bytes, extension: output.extension };
+    const response = await fetch(output.url);
+    if (!response.ok)
+        throw new Error(`Failed to download generated ${media}: ${response.status} ${response.statusText}`);
+    const bytes = Buffer.from(await response.arrayBuffer());
+    if (bytes.length === 0)
+        throw new Error(`Generated ${media} download was empty`);
+    const extension = extensionFromUrl(output.url, media)
+        ?? extensionFromContentType(response.headers.get("content-type"), media)
+        ?? FALLBACK_EXTENSION[media];
+    return { bytes, extension, url: output.url };
+}
+function provenance(meta, file) {
+    return {
+        model: meta.model,
+        prompt: meta.prompt,
+        source_images: meta.sourceImageIds,
+        source_videos: meta.sourceVideoIds,
+        source_audios: meta.sourceAudioIds,
+        ...(file.url ? { download_url: file.url } : {}),
+        download_url_content_hash: contentHash(file.bytes),
+        generated_at: new Date().toISOString(),
+    };
+}
+async function writeOriginal(engine, assetId, filename, bytes) {
+    const written = await engine.files.write(assetId, filename, bytes);
+    if (!written.ok)
+        throw new Error(`Failed to write ${filename}: ${written.error.message}`);
+}
+async function completeImage(completion, file) {
+    const extension = file.extension === "jpeg" ? "jpg" : file.extension;
+    await writeOriginal(completion.engine, completion.assetId, `original.${extension}`, file.bytes);
+    await queue.enqueue(`process-image-${completion.assetId}-${Date.now()}`, {
+        type: "process_image",
+        assetId: completion.assetId,
+        bookId: completion.bookId,
+        extension: `.${extension}`,
+        provenance: provenance(completion.meta, file),
+    });
+}
+async function enqueueReplaceAudio(completion, videoPath, origin) {
+    const replace = record(completion.meta.replaceAudioWith);
+    if (!replace || typeof replace.assetId !== "string" || typeof replace.filename !== "string")
+        return false;
+    const manifest = await completion.engine.files.manifest(replace.assetId);
+    if (!manifest.ok)
+        throw new Error(`replaceAudioWith asset not found: ${manifest.error.message}`);
+    await queue.enqueue(`replace-audio-${completion.assetId}-${Date.now()}`, {
+        type: "replace_video_audio",
+        inputPath: videoPath,
+        audioPath: join(manifest.value.path, replace.filename),
+        outputPath: join(completion.assetDir, "original_audio_replaced.mp4"),
+        assetId: completion.assetId,
+        assetDir: completion.assetDir,
+        audioAssetId: `${replace.assetId}/${replace.filename}`,
+        bookId: completion.bookId,
+        extraProvenance: origin,
+    });
+    return true;
+}
+async function completeVideo(completion, file) {
+    const filename = `original.${file.extension}`;
+    await writeOriginal(completion.engine, completion.assetId, filename, file.bytes);
+    const videoPath = join(completion.assetDir, filename);
+    const origin = provenance(completion.meta, file);
+    if (await enqueueReplaceAudio(completion, videoPath, origin))
+        return;
+    await queue.enqueue(`process-video-${completion.assetId}-${Date.now()}`, {
+        type: "process_video",
+        videoPath,
+        assetId: completion.assetId,
+        assetDir: completion.assetDir,
+        bookId: completion.bookId,
+        provenance: origin,
+    });
+}
+async function toMp3(file) {
+    if (file.extension === "mp3")
+        return file.bytes;
+    const dir = await mkdtemp(join(tmpdir(), "videobook-media-audio-"));
+    try {
+        const input = join(dir, `input.${file.extension}`);
+        const output = join(dir, "output.mp3");
+        await writeFile(input, file.bytes);
+        await runFfmpeg(["-y", "-i", input, "-c:a", "libmp3lame", "-q:a", "2", output]);
+        return await readFile(output);
+    }
+    finally {
+        await rm(dir, { recursive: true, force: true });
+    }
+}
+async function probeDuration(path) {
+    try {
+        const probe = await probeVideo(path);
+        return Number.isFinite(probe.duration) && probe.duration > 0 ? probe.duration : undefined;
+    }
+    catch {
+        return undefined;
+    }
+}
+async function completeAudio(completion, file) {
+    await writeOriginal(completion.engine, completion.assetId, "original.mp3", await toMp3(file));
+    const duration = await probeDuration(join(completion.assetDir, "original.mp3"));
+    const metadata = {
+        ...provenance(completion.meta, file),
+        origin: "generate_audio",
+        ...(duration !== undefined ? { duration } : {}),
+    };
+    const written = await completion.engine.metadata.artifacts.write(completion.assetId, "original", metadata);
+    if (!written.ok)
+        throw new Error(`Failed to write artifact metadata: ${written.error.message}`);
+}
+export async function completeMediaGeneration(completion) {
+    const media = mediaOf(completion.meta);
+    const output = completion.outputs[0];
+    if (!output)
+        throw new Error(`Provider returned no ${media}`);
+    const file = await download(output, media);
+    if (media === "image")
+        return completeImage(completion, file);
+    if (media === "video")
+        return completeVideo(completion, file);
+    return completeAudio(completion, file);
+}
+const URL_KEYS = ["video", "image", "audio", "audio_file", "images", "videos", "audios"];
+function urlsIn(value) {
+    if (typeof value === "string")
+        return /^https?:\/\//u.test(value) ? [value] : [];
+    if (Array.isArray(value))
+        return value.flatMap(urlsIn);
+    const object = record(value);
+    if (!object)
+        return [];
+    if (typeof object.url === "string")
+        return [object.url];
+    return URL_KEYS.flatMap((key) => urlsIn(object[key]));
+}
+export function providerResultUrls(resultUrls, resultObject) {
+    if (resultUrls.length > 0)
+        return [...resultUrls];
+    const object = record(resultObject);
+    if (!object)
+        return [];
+    const direct = URL_KEYS.flatMap((key) => urlsIn(object[key]));
+    if (direct.length > 0)
+        return direct;
+    for (const key of ["video_url", "image_url", "audio_url", "resultUrls"]) {
+        const found = urlsIn(object[key]);
+        if (found.length > 0)
+            return found;
+    }
+    return [];
+}

package/dist/media/cost.js

@@ -0,0 +1,45 @@
+import { computeEstimate } from "@apicity/cost";
+import { buildWireRequest } from "./request.js";
+function costProvider(model) {
+    return model.provider;
+}
+function costPayload(model, request, input) {
+    return model.provider === "fal" ? input : request;
+}
+function numeric(value) {
+    if (typeof value === "number" && Number.isFinite(value))
+        return value;
+    if (typeof value === "string" && value.trim() !== "" && Number.isFinite(Number(value)))
+        return Number(value);
+    return undefined;
+}
+function durationHint(input) {
+    const seconds = numeric(input.duration) ?? numeric(input.duration_seconds)
+        ?? (numeric(input.music_length_ms) !== undefined ? numeric(input.music_length_ms) / 1000 : undefined);
+    return seconds !== undefined && seconds > 0 ? { durationSeconds: seconds } : {};
+}
+function combinedHints(input, hints) {
+    const combined = { ...durationHint(input), ...hints };
+    return combined.durationSeconds === undefined && combined.inputDurationSeconds !== undefined
+        ? { ...combined, durationSeconds: combined.inputDurationSeconds }
+        : combined;
+}
+function estimateWireRequest(model, request, input, hints = {}) {
+    return computeEstimate({
+        provider: costProvider(model),
+        payload: costPayload(model, request, input),
+        endpoint: model.apiId,
+        costHints: combinedHints(input, hints),
+    });
+}
+export function estimateGeneration(model, call, hints = {}) {
+    const built = buildWireRequest(model, call);
+    if (!built.ok)
+        return built;
+    try {
+        return { ok: true, estimate: estimateWireRequest(model, built.request, built.input, hints), request: built.request };
+    }
+    catch (error) {
+        return { ok: false, error: error instanceof Error ? error.message : String(error) };
+    }
+}