motionrender 0.2.3 → 0.2.5

package/dist/cli.js

@@ -2951,25 +2951,23 @@ function inferChatTask(messages2, options3) {
   return "compose";
 }
 function createProvider(config = {}) {
-  const id2 = (config.provider || process.env.OVE_AI_PROVIDER || (process.env.FREELLMAPI_BASE_URL ? "freellmapi" : "") || (process.env.NVIDIA_API_KEY ? "nvidia" : process.env.GROQ_API_KEY ? "groq" : process.env.ANTHROPIC_API_KEY ? "anthropic" : "")).toLowerCase();
+  const id2 = (config.provider || process.env.OVE_AI_PROVIDER || (process.env.FREELLMAPI_BASE_URL || process.env.FREELLMAPI_API_KEY ? "freellmapi" : "") || (process.env.NVIDIA_API_KEY ? "nvidia" : process.env.GROQ_API_KEY ? "groq" : process.env.ANTHROPIC_API_KEY ? "anthropic" : "") || "freellmapi").toLowerCase();
   switch (id2) {
+    case "mock":
+      return new MockProvider();
     case "freellmapi":
     case "openai":
-      return attachGroqAudio(new FreeLlmApiProvider(config), config);
+      return new FreeLlmApiProvider(config);
+    case "groq":
+      return new GroqProvider(config);
     case "nvidia":
     case "nim":
       return attachGroqAudio(new NvidiaProvider(config), config);
     case "anthropic":
     case "claude":
       return attachGroqAudio(new AnthropicProvider(config), config);
-    case "groq":
-      return new GroqProvider(config);
-    case "mock":
-      return new MockProvider();
-    case "":
-      throw new Error("No AI provider chosen: pass --provider freellmapi|nvidia|groq|anthropic|mock, or set FREELLMAPI_BASE_URL, NVIDIA_API_KEY, GROQ_API_KEY or ANTHROPIC_API_KEY");
     default:
-      throw new Error(`Unknown AI provider "${id2}"; use freellmapi, nvidia, groq, anthropic or mock`);
+      return new FreeLlmApiProvider(config);
   }
 }
 function describeProvider(provider) {
@@ -3259,6 +3257,65 @@ var init_providers = __esm({
         if (!text) throw new Error("FreeLLMAPI chat returned no text. Check active models and upstream keys in your FreeLLMAPI dashboard.");
         return text;
       }
+      async speak(text, options3 = {}) {
+        const response = await request(`${this.baseUrl}/audio/speech`, {
+          method: "POST",
+          signal: timeoutSignal(options3.signal),
+          headers: jsonHeaders(this.headers()),
+          body: JSON.stringify({
+            model: "auto",
+            input: text,
+            voice: options3.voice ?? "alloy"
+          })
+        }, "FreeLLMAPI speech");
+        return { data: new Uint8Array(await response.arrayBuffer()), extension: "mp3" };
+      }
+      async transcribe(audio, filename, options3 = {}) {
+        const form = new FormData();
+        form.append("file", new Blob([audio]), filename || "audio.mp3");
+        form.append("model", "auto");
+        form.append("response_format", "verbose_json");
+        if (options3.language) form.append("language", options3.language);
+        const response = await request(`${this.baseUrl}/audio/transcriptions`, {
+          method: "POST",
+          signal: timeoutSignal(options3.signal),
+          headers: this.headers(),
+          body: form
+        }, "FreeLLMAPI transcription");
+        const data = await response.json();
+        return {
+          text: String(data?.text ?? ""),
+          words: Array.isArray(data?.words) ? data.words.map((w) => ({
+            text: String(w.word ?? w.text ?? ""),
+            start: Number(w.start) || 0,
+            end: Number(w.end) || 0
+          })) : [],
+          cues: Array.isArray(data?.segments) ? data.segments.map((s) => ({
+            text: String(s.text ?? ""),
+            start: Number(s.start) || 0,
+            end: Number(s.end) || 0
+          })) : []
+        };
+      }
+      async image(prompt, options3 = {}) {
+        const size = options3.size ?? "1024x1024";
+        const response = await request(`${this.baseUrl}/images/generations`, {
+          method: "POST",
+          signal: timeoutSignal(options3.signal),
+          headers: jsonHeaders(this.headers()),
+          body: JSON.stringify({
+            model: "auto",
+            prompt,
+            n: 1,
+            size,
+            response_format: "b64_json"
+          })
+        }, "FreeLLMAPI image");
+        const data = await response.json();
+        const b64 = data?.data?.[0]?.b64_json;
+        if (!b64) throw new Error("FreeLLMAPI did not return image data");
+        return { data: Buffer.from(b64, "base64"), extension: "png" };
+      }
     };
     attachGroqAudio = (provider, config) => {
       if (provider.speak || provider.transcribe || !process.env.GROQ_API_KEY) return provider;
@@ -4430,6 +4487,13 @@ var init_template = __esm({
     };
     templatesFor = (templates, size) => templates.filter((template) => sameShape(template, size));
     systemText = `You fill in ready-made video templates. Reply with JSON only: {"template": "<id>", "theme": "<theme id, only for templates that take a theme>", "values": {"<variable>": <value>, ...}}.
+Choose the template that best matches the story structure and topic of the brief (stories and stats -> story-tall; features and cards -> showcase-tall; tips and steps -> social-reel).
+For templates that take a theme, ALWAYS choose the theme that best matches the mood, industry, and visual tone of the brief:
+- "midnight": modern tech, SaaS, B2B, cloud, finance
+- "neon": gaming, nightlife, social apps, crypto, cutting-edge AI
+- "espresso": artisan craft, food & beverage, coffee, lifestyle, podcasts, warmth
+- "paper": clean editorial, education, publishing, thought leadership, light minimalist
+- "signal": high-impact, fitness, sports, breaking news, urgency
 Picture variables (image, image2\u2026) take the id of one supplied picture, or "" for none; never invent ids. Examples only show the length and kind of text wanted: never reuse their wording (including labels like INTRODUCING); use the brief's own names, numbers and details, and prefer a concrete line over a generic one such as "Ready to start?". Rules: write the words viewers will read, never a description of the video; do not invent dates, prices or claims that are not in the brief; use only the listed variable names; write real copy for the brief, about as long as each example (never much longer); "color" values are hex like "#f4d58d"; list values are arrays of objects with the listed fields.
 If no template suits the brief, reply {"template": "none"}.`;
   }
@@ -4514,7 +4578,7 @@ function narrationScene(composition, options3) {
   return next;
 }
 async function narrate(provider, options3) {
-  if (!provider.speak) throw new Error(`The ${provider.id} provider cannot make speech; use --provider groq (or add a speak() adapter)`);
+  if (!provider.speak) throw new Error(`The ${provider.id} provider cannot make speech (or add a speak() adapter)`);
   const brief = options3.brief?.trim();
   let text = options3.text?.trim() ?? "";
   if (!text && brief) {
@@ -4561,7 +4625,7 @@ async function narrate(provider, options3) {
   return { audio: audioFile, captions: captionsFile, durationSeconds, words: transcript.words.length, composition, text };
 }
 async function captionAudio(provider, options3) {
-  if (!provider.transcribe) throw new Error(`The ${provider.id} provider cannot transcribe; use --provider groq (or add a transcribe() adapter)`);
+  if (!provider.transcribe) throw new Error(`The ${provider.id} provider cannot transcribe (or add a transcribe() adapter)`);
   const audioFile = resolve6(options3.audio);
   const data = await readFile8(audioFile);
   options3.onProgress?.("Transcribing\u2026");
@@ -4812,7 +4876,7 @@ function editComposition(options3) {
   return composition;
 }
 async function editVideo(provider, options3) {
-  if (!provider.transcribe) throw new Error(`The ${provider.id} provider cannot transcribe; use --provider groq (or add a transcribe() adapter)`);
+  if (!provider.transcribe) throw new Error(`The ${provider.id} provider cannot transcribe (or add a transcribe() adapter)`);
   const clip2 = resolve7(options3.clip);
   const outDir = resolve7(options3.outDir);
   const width = options3.width ?? 1080, height = options3.height ?? 1920, fps = options3.fps ?? 30;
@@ -4920,7 +4984,7 @@ function shortsCandidatePlan(candidate, cut, maxSilence, keepFillers) {
   return cut ? planCuts(slice, { maxSilence, keepFillers, padding: 0.2 }) : planCuts(slice, { maxSilence: Number.MAX_SAFE_INTEGER, keepFillers: true, padding: 0.2 });
 }
 async function shortsFrom(provider, options3) {
-  if (!provider.transcribe) throw new Error(`The ${provider.id} provider cannot transcribe; use --provider groq (or add a transcribe() adapter)`);
+  if (!provider.transcribe) throw new Error(`The ${provider.id} provider cannot transcribe (or add a transcribe() adapter)`);
   const clip2 = resolve8(options3.clip);
   const outDir = resolve8(options3.outDir);
   const width = options3.width ?? 1080, height = options3.height ?? 1920, fps = options3.fps ?? 30;
@@ -5048,7 +5112,7 @@ function imageSize(size) {
   return sizes[size.toLowerCase()] ?? size;
 }
 async function generateImage(provider, options3) {
-  if (!provider.image) throw new Error(`The ${provider.id} provider cannot generate images; use --provider mock (Groq has no image model; or add an image() adapter)`);
+  if (!provider.image) throw new Error(`The ${provider.id} provider cannot generate images (or add an image() adapter)`);
   const generated = await provider.image(options3.prompt, { size: imageSize(options3.size), signal: options3.signal });
   const file2 = resolve9(/\.(png|jpe?g|webp)$/i.test(options3.out) ? options3.out : `${options3.out}.${generated.extension}`);
   await mkdir12(dirname9(file2), { recursive: true });
@@ -5970,17 +6034,31 @@ var init_plan = __esm({
     init_registry();
     paths = ["template", "json", "react"];
     allowedPaths = (options3) => (options3.paths ?? paths).filter((path) => path !== "react" || options3.allowReact);
-    systemText2 = (options3) => `You are the production planner for Open Video Engine. Decide how the brief below should be produced, before anything is generated.
+    systemText2 = (options3) => `You are the creative production planner for Open Video Engine. Decide how the brief below should be produced, before anything is generated.
 Reply with ONE JSON object and nothing else:
 {
   "path": "template" | "json" | "react",
-  "template": "<template id>",            // only with "template": the one template that suits the brief
+  "template": "<template id>",            // only with "template": the one template that best suits the brief's structure and topic
   "theme": "<theme id>",                  // optional, only with a template that takes a theme
   "idea": "<what the video says, in one line>",
   "beats": [{ "name": "Hook", "onScreen": "what is on screen, concretely", "seconds": 3, "motion": "how it moves in" }],
   "reasons": ["why this path suits this brief"]
 }
-Templates are the primary path: choose "template" whenever one listed template suits the brief's kind and aspect ratio, even when its length differs \u2014 a template's length is its own, and a length difference is not a reason to leave the template route. Choose "json" only when no template fits the kind and aspect: a bespoke composition with scene layers, camera moves and kinetic type. Choose "react" only when allowed: it writes React code for bespoke high-end motion graphics. Every beat needs a concrete picture, not a mood.
+Templates provide proven layouts, but pick the template that specifically fits the brief's narrative style:
+- Testimonials, customer stories, quotes, metrics, or personal founder stories: choose "story-tall" or "quote-card".
+- Multi-feature tours, app UI cards, feature lists, and product showcases: choose "showcase-tall", "product-wide" or "product-card".
+- Quick tips, 3-step processes, how-to guides, and numbered reels: choose "social-reel".
+- Company keynotes, presentations, and product launch films: choose "keynote-wide", "launch-wide" or "case-study".
+- Choose "json" when the brief requires a bespoke kinetic video with custom motion graphics, camera moves, and scene architecture.
+- Choose "react" only when allowed: it writes React code for bespoke high-end motion graphics.
+
+Themes: For templates that take a theme, ALWAYS choose the theme matching the tone and brand of the brief:
+- "midnight": modern tech, SaaS, B2B, cloud, finance (deep slate navy, sky blue).
+- "neon": gaming, nightlife, social apps, crypto, cutting-edge AI (dark violet, hot magenta, electric lime).
+- "espresso": artisan craft, food & beverage, coffee, lifestyle, podcasts, warmth (dark espresso, amber gold).
+- "paper": clean editorial, education, publishing, thought leadership (clean light aesthetic, deep ink).
+- "signal": high-impact, fitness, sports, breaking news, urgency (bold monochrome with alert red).
+Every beat needs a concrete picture, not a mood.
 Reply with "path" set to one of: ${allowedPaths(options3).join(" | ")}.
 Available capabilities:
 ${describeCapabilities(options3.registry, options3)}`;
@@ -6115,7 +6193,7 @@ async function cliRegistry() {
 async function composeSmart(provider, options3, mode = {}) {
   const needsFull = mode.full || !!options3.fields?.length || (options3.assets ?? []).some((asset) => asset.type !== "image");
   let plan;
-  if (!mode.noPlan) {
+  if (!mode.noPlan && !mode.full && !mode.template) {
     const registry = mode.registry ?? await cliRegistry();
     try {
       plan = await planBuild(provider, { ...options3, registry, maxAttempts: Math.min(options3.maxAttempts ?? 2, 2) });
@@ -6123,13 +6201,16 @@ async function composeSmart(provider, options3, mode = {}) {
       options3.onProgress?.(`Planning skipped (${error.message.split("\n")[0]}); using the built-in routing.`);
     }
   }
-  const templatePath = plan ? plan.path === "template" : !needsFull;
+  const templatePath = (plan ? plan.path === "template" : !needsFull) || Boolean(mode.template);
   if (templatePath && !needsFull) {
     try {
       const choices = await templateChoices();
-      const picked = plan?.template ? choices.filter((choice) => choice.id === plan.template) : choices;
-      const themes = plan?.theme ? (await listThemes()).filter((theme) => theme.id === plan.theme) : await listThemes();
-      const filled2 = await composeFromTemplate(provider, picked.length ? picked : choices, { ...options3, themes });
+      const targetTemplateId = mode.template ?? plan?.template;
+      const picked = targetTemplateId ? choices.filter((choice) => choice.id === targetTemplateId) : choices;
+      const targetThemeId = mode.theme ?? plan?.theme;
+      const allThemes = await listThemes();
+      const themes = targetThemeId ? allThemes.filter((theme) => theme.id === targetThemeId) : allThemes;
+      const filled2 = await composeFromTemplate(provider, picked.length ? picked : choices, { ...options3, themes: themes.length ? themes : allThemes });
       if (filled2) {
         const wanted = options3.durationSeconds;
         const current = filled2.composition.durationFrames / filled2.composition.fps;
@@ -7188,7 +7269,9 @@ function createHandlers(env) {
         maxAttempts: integerIn(input.maxAttempts, 3, "maxAttempts", 1, 10),
         onProgress: (message2) => progress(message2)
       };
-      const result = await capabilities.composeSmart(ai, options3, { full });
+      const theme = typeof input.theme === "string" && input.theme.trim() ? input.theme.trim() : void 0;
+      const template = typeof input.template === "string" && input.template.trim() ? input.template.trim() : void 0;
+      const result = await capabilities.composeSmart(ai, options3, { full, theme, template });
       let composition = result.composition;
       let narration;
       if (options3.narration) {

package/dist/types/packages/ai/src/providers.d.ts

@@ -130,6 +130,12 @@ export declare class FreeLlmApiProvider implements AiProvider {
     private headers;
     resolveModel(task: ChatTask, explicitModel?: string, signal?: AbortSignal): Promise<string>;
     chat(messages: ChatMessage[], options?: ChatOptions): Promise<string>;
+    speak(text: string, options?: SpeakOptions): Promise<AudioResult>;
+    transcribe(audio: Uint8Array, filename: string, options?: {
+        language?: string;
+        signal?: AbortSignal;
+    }): Promise<Transcript>;
+    image(prompt: string, options?: ImageOptions): Promise<ImageResult>;
 }
 export declare class MockProvider implements AiProvider {
     id: string;

package/dist/types/packages/cli/src/templates.d.ts

@@ -56,6 +56,8 @@ export declare function composeSmart(provider: AiProvider, options: ComposeOptio
     full?: boolean;
     noPlan?: boolean;
     registry?: Registry;
+    theme?: string;
+    template?: string;
 }): Promise<ComposeResult & {
     template?: string;
     source?: Record<string, unknown>;

package/dist/types/packages/jobs/src/env.d.ts

@@ -30,6 +30,9 @@ export interface JobPaths {
 export type JobCapabilities = {
     composeSmart(provider: AiProvider, options: ComposeOptions, mode: {
         full?: boolean;
+        noPlan?: boolean;
+        theme?: string;
+        template?: string;
     }): Promise<ComposeResult & {
         template?: string;
         source?: Record<string, unknown>;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "motionrender",
-  "version": "0.2.3",
+  "version": "0.2.5",
   "description": "JSON-first, frame-deterministic motion graphics video engine: render compositions to MP4, WebM, ProRes, GIF or PNG frames from Node or the command line.",
   "license": "MIT",
   "type": "module",
@@ -77,4 +77,4 @@
       "optional": true
     }
   }
-}
+}