@kolbo/mcp 1.76.3 → 1.76.5

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@kolbo/mcp",
-  "version": "1.76.3",
+  "version": "1.76.5",
   "description": "Kolbo AI MCP Server - Generate images, videos, music, speech, and sound effects from Claude Code",
   "main": "src/index.js",
   "bin": {

package/skill/references/models/seedance.md

@@ -110,6 +110,7 @@ These elevate rich cinematic / reference-anchored sequences. For a short, tight,
 ## Dialogue & expression
 
 - Dialogue goes in quotes and may be in ANY language (Hebrew included). For silent tension, deliver it as expression, not speech: `He does not speak. His expression clearly says: "…"`.
+- **Seedance PERFORMS quoted dialogue natively** — synced voices, lip movement, and room tone come out of the video model itself. Never route scene dialogue through TTS (`generate_speech`) or `generate_lipsync`; write each line in quotes inside its shot beat (`DANIEL says: "…"`) and generate once.
 
 ## Content tone
 

package/skill/references/models/seedance25.md

@@ -9,7 +9,7 @@ Load this file when the user wants a **Seedance 2.5** video (they said "2.5" / "
 
 **Kolbo MCP routing:** `generate_video` or `generate_elements` (refs / Visual DNA / first-last). Run `list_models({ type: "text_to_video" })` and pick the Seedance 2.5 variant by name.
 
-**Audio:** Seedance 2.5 still emits real synced audio. `list_models` may show `sound_generation_type: none` because there is no in-app toggle (`sound_baked_in: true`). Do not tell the user the model is silent.
+**Audio:** Seedance 2.5 still emits real synced audio. `list_models` may show `sound_generation_type: none` because there is no in-app toggle (`sound_baked_in: true`). Do not tell the user the model is silent. Quoted dialogue in the prompt is PERFORMED — synced voices, lip movement, room tone — so scene dialogue never goes through `generate_speech` or `generate_lipsync`; write the lines in quotes inside their shot beats.
 
 ## What's NEW in 2.5 (verified — never hedge)
 

package/src/apps/widgets/generation.js

@@ -668,15 +668,21 @@ function renderAudio(sc, urls) {
     var titleBase = track.title || sc.title || (TOOL_TITLES[sc.tool] || 'Audio');
     var title = titleBase + (urls.length > 1 ? ' — Track ' + (i + 1) : '');
     var duration = track.duration != null ? track.duration : sc.duration;
-    var artwork = track.thumbnail_url || sc.thumbnail_url;
+    // The voice's own portrait is the artwork for speech — a generic note glyph
+    // told the user nothing about the one thing that defines the take.
+    var artwork = track.thumbnail_url || sc.thumbnail_url || sc.voice_thumbnail;
+    var placeholder = sc.tool === 'generate_speech' ? ICONS.mic : ICONS.audio;
     return '<div class="k-audio-row k-generated-audio">' +
-      (artwork ? '<img class="k-audio-art" src="' + esc(artwork) + '" alt="" loading="lazy">' :
-        '<div class="k-audio-art k-audio-placeholder">' + ICONS.audio + '</div>') +
+      (artwork ? '<img class="k-audio-art" src="' + esc(artwork) + '" alt="" loading="lazy" onerror="this.style.display=\\'none\\'">' :
+        '<div class="k-audio-art k-audio-placeholder">' + placeholder + '</div>') +
       '<div class="k-audio-meta"><div class="k-audio-title">' + esc(title) + '</div>' +
-      // Resolved name first: a per-track model field is the raw id, and every
-      // track in one generation came from the same model anyway.
-      '<div class="k-audio-sub">' + esc(modelLabel(sc) || track.model || '') +
-      (duration ? ' · ' + fmtDur(duration) : '') + '</div></div>' +
+      // Voice FIRST where there is one — on a speech row, who is speaking is the
+      // thing that defines the take; the engine is secondary. Resolved names
+      // only: a per-track model field is the raw id, and every track in one
+      // generation came from the same model anyway.
+      '<div class="k-audio-sub">' +
+      [voiceLabel(sc), modelLabel(sc) || track.model, duration ? fmtDur(duration) : '']
+        .filter(Boolean).map(esc).join(' · ') + '</div></div>' +
       '<button class="k-btn k-audio-download" data-audio-download="' + esc(u) +
       '" aria-label="Download ' + esc(title) + '">' + ICONS.download + ' Download</button>' +
       '<audio class="k-audio-player" src="' + esc(u) + '" controls preload="none" aria-label="Play ' +

package/src/tools/_shared.js

@@ -611,6 +611,7 @@ async function uiGenerating(p) {
       ? { failed_submissions: p.failed_submissions } : {}),
     ...(p.warning ? { _warning: p.warning } : {}),
     _widget_note: 'A live Kolbo widget is rendering this generation for the user (progress + final result + action buttons). Tell the user it is generating and the card above will update — do NOT poll in a loop. If you need the output URLs (e.g. for a follow-up edit or a report), call get_generation_status ONCE with wait=true — it blocks until done. Tracking several generations? Pass ALL their ids in generation_ids in that one call.',
+    _paid_note: 'This generation is RUNNING and the user is paying for it. If you now realize the tool, model, or parameters were wrong, call cancel_generation with this generation_id FIRST, then start the replacement — never leave a wrong generation running alongside its retry (the user gets two cards and two charges).',
   }, null, 2);
   return uiResult(UI.generation, text, structured);
 }
@@ -659,6 +660,14 @@ async function uiCompleted(p, textPayload, extraContent) {
     // above which assume everything finished together. Only set when the
     // caller actually has this shape; every existing caller is unaffected.
     ...(Array.isArray(p.items) ? { items: p.items } : {}),
+    // The voice, by name and portrait. uiGenerating has carried this since the
+    // chips were introduced; uiCompleted never did, so it silently dropped a
+    // resolved voice its caller had already looked up — every FINISHED speech
+    // card fell back to `settings.voice`, printing a raw ElevenLabs id where the
+    // name belongs and rendering the generic note placeholder instead of the
+    // voice's face. Exactly the "never show a raw id on a card" rule, broken on
+    // the one path the user actually ends up looking at.
+    ...(p.voice ? { voice_name: p.voice.name, voice_thumbnail: p.voice.thumbnail } : {}),
     // The RAW generation state, when the caller has one. `phase` above is
     // hardcoded 'completed' — it means "this tool call finished", not "the
     // generation finished" — so a status check on a still-running job looked

package/src/tools/generate.js

@@ -541,7 +541,7 @@ function registerGenerateTools(server, client, options = {}) {
   // retired textToVideoGeneration path and was stale.
   server.tool(
     'generate_video',
-    'Generate a video from a text prompt using Kolbo AI. For SEVERAL different videos, pass all their prompts in `prompts` in ONE call (one combined widget) — never a series of separate calls. For animating an existing still image into motion, use generate_video_from_image instead. For a coordinated multi-scene video campaign, use generate_creative_director with workflow_type="video". Supports reference images (for style/composition guidance) and Visual DNA for character consistency. Returns the final video URL when complete.',
+    'Generate a video from a text prompt using Kolbo AI. For SEVERAL different videos, pass all their prompts in `prompts` in ONE call (one combined widget) — never a series of separate calls. For animating an existing still image into motion, use generate_video_from_image instead. For a coordinated multi-scene video campaign, use generate_creative_director with workflow_type="video". Supports reference images (for style/composition guidance) and Visual DNA for character consistency. ROUTE BEFORE CALLING: when reference images anchor IDENTITY (specific characters, a specific product, a location that must match) — especially 2+ of them — that is generate_elements, not this tool; reference_images here are loose style/composition hints. Decide the right tool FIRST: a mis-routed call still starts a PAID generation, and switching tools afterwards without cancel_generation leaves the user paying for both. Returns the final video URL when complete.',
     {
       prompt: z.string().optional().describe('Text description of the video to generate. Required unless `prompts` is provided.'),
       prompts: promptsField('videos'),
@@ -855,11 +855,19 @@ function registerGenerateTools(server, client, options = {}) {
       if (poll.timedOut) return poll.timedOut;
       const result = poll.result;
 
+      // What ACTUALLY ran, not what was asked for. The status result reports the
+      // engine's own choice (`model: "google_tts"`, `voice: "he-IL-Chirp3-HD-Kore"`)
+      // and for an omitted model that is the ONLY place the real answer appears —
+      // the card was labelling those "Smart Select", which is not even a text-to-
+      // speech option, and naming the voice the caller typed rather than the one
+      // that spoke. Same resolution addDisplayNames does for the polling path.
+      const ranVoice = (await voiceInfo(client, result.result?.voice).catch(() => null)) || voiceRecord;
       return uiCompleted({
-        tool: 'generate_speech', kind: 'audio', gen, client, model, prompt: text,
-        voice: voiceRecord,
+        tool: 'generate_speech', kind: 'audio', gen, client, prompt: text,
+        model: result.result?.model || model,
+        voice: ranVoice,
         settings: {
-          voice: voice || 'Rachel',
+          voice: (ranVoice && ranVoice.name) || voice || 'Rachel',
           style: selected_style || emotion || style_instructions_preset_id || style_instructions,
           speaking_speed,
           language,