@kolbo/mcp 1.110.3 → 1.111.0

package/README.md

@@ -147,7 +147,7 @@ Without the optional skill, the config block alone already exposes every tool
 | `generate_lipsync` | Source image/video + audio → lipsynced video (Sync-3 adds active-speaker selection, emotion, model mode, temperature) |
 | `generate_creative_director` | One brief → N coordinated scenes (image or video) |
 | `generate_music` | Text (+ optional lyrics) → song. Style, title, negative tags, length, and Suno fine-controls (style weight, weirdness, audio weight, persona / singing voice) |
-| `generate_speech` | Text + voice → spoken audio. Full expressive/style control: Google/Gemini named voice-direction presets (`style_instructions_preset_id`: warm/dramatic/whisper/…) or free-form `style_instructions`, preset styles + emotions (DeepDub / MiniMax / Cartesia), speed, accent/`language`, and per-provider voice settings (ElevenLabs similarity/style, DeepDub accent/variance/tempo, MiniMax pitch/volume/intensity/timbre). Status returns the same fields for reuse. |
+| `generate_speech` | Text + voice → spoken audio. Full expressive/style control: Google/Gemini named voice-direction presets (`style_instructions_preset_id`: warm/dramatic/whisper/…) or free-form `style_instructions`, preset styles + emotions (DeepDub / MiniMax / Cartesia), speed, accent/`language`, and per-provider voice settings (ElevenLabs stability/similarity/style/speaker boost/seed, `enforce_language`, `output_format` mp3/wav, `text_normalization`; DeepDub accent/variance/tempo, MiniMax pitch/volume/intensity/timbre), plus `audio_effect` (phone/radio/megaphone/small_room/hall/distant) + `audio_effect_strength` for any voice at no extra credits. Status returns the same fields for reuse. |
 | `generate_sound` | Text → sound effect. Duration, prompt influence, and per-provider controls (Stable Audio guidance, Kie loop/tempo/key, Seed-Audio voice/speed/volume/pitch + reference audio/image) |
 | `generate_3d` | Text or reference images → 3D model (GLB/FBX/OBJ/USDZ) |
 | `transcribe_audio` | Audio/video URL or file → text + SRT subtitles. Language, speaker diarization, audio-event tagging, and SRT formatting (words/line, lines/subtitle, caption stretch) |

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@kolbo/mcp",
-  "version": "1.110.3",
+  "version": "1.111.0",
   "description": "Kolbo AI MCP Server - Generate images, videos, music, speech, and sound effects from Claude Code",
   "main": "src/index.js",
   "bin": {

package/src/apps/widgets/generation.js

@@ -230,6 +230,45 @@ function modelLabel(sc) {
 }
 function voiceLabel(sc) { return sc.voice_name || sc.voice || (sc.settings || {}).voice; }
 
+// generate_speech controls. What the API reports it APPLIED wins: \`speech\`
+// (status / completed payload; a key there is authoritative even when null),
+// then the raw status result's own fields, then what the caller submitted.
+var SPEECH_KEYS = ['turbo', 'speaking_speed', 'stability', 'similarity_boost', 'enforce_language',
+  'output_format', 'audio_effect', 'audio_effect_strength', 'seed'];
+var SPEECH_FORMATS = { mp3_44100_128: 'MP3 128', mp3_44100_192: 'MP3 192', wav_44100: 'WAV' };
+function withSpeech(settings, a) {
+  SPEECH_KEYS.forEach(function (k) { if (a && a[k] != null) settings[k] = a[k]; });
+  return settings;
+}
+function speechVal(sc, k) {
+  if (sc.speech && Object.prototype.hasOwnProperty.call(sc.speech, k)) return sc.speech[k];
+  var vs = sc.voice_settings || {};
+  if (vs[k] != null) return vs[k];
+  if (sc[k] != null) return sc[k];
+  return (sc.settings || {})[k];
+}
+function num(v) { return String(Math.round(Number(v) * 100) / 100); }
+function speechChipsHTML(sc) {
+  if (sc.tool !== 'generate_speech' && !sc.speech) return '';
+  var h = '', v;
+  if (speechVal(sc, 'turbo') === true) h += chip('Turbo');
+  v = speechVal(sc, 'speaking_speed'); if (v != null) h += chip('Speed ' + esc(num(v)) + '×');
+  v = speechVal(sc, 'stability'); if (v != null) h += chip('Stability ' + esc(num(v)));
+  v = speechVal(sc, 'similarity_boost'); if (v != null) h += chip('Similarity ' + esc(num(v)));
+  v = speechVal(sc, 'enforce_language');
+  h += chip('Language ' + (v ? esc(String(v).toUpperCase()) : 'Auto'));
+  v = speechVal(sc, 'output_format'); if (v) h += chip(esc(SPEECH_FORMATS[v] || v));
+  v = speechVal(sc, 'audio_effect');
+  if (v && v !== 'none') {
+    var strength = speechVal(sc, 'audio_effect_strength');
+    var name = String(v).replace(/_/g, ' ');
+    h += chip(esc(name.charAt(0).toUpperCase() + name.slice(1)) +
+      (strength != null && Number(strength) < 1 ? ' ' + Math.round(Number(strength) * 100) + '%' : ''));
+  }
+  v = speechVal(sc, 'seed'); if (v != null) h += chip('Seed ' + esc(String(v)));
+  return h;
+}
+
 // @VisualDNA / #Moodboard mentions are the tag syntax the server resolves into
 // real reference assets — rendering them as flat prose hid the single most
 // consequential part of the prompt. Escape FIRST, then wrap: the pattern only
@@ -340,6 +379,7 @@ function renderChips(sc) {
       : ICONS.mic;
     h += chip(face + ' ' + esc(voice));
   }
+  h += speechChipsHTML(sc);
   if (s.mode) h += chip(esc(editLabel(s.mode)));
   if (Array.isArray(s.details)) s.details.forEach(function (d) { if (d) h += chip(esc(String(d))); });
   if (many && !kind) h += chip(sc.count + ' items');
@@ -1358,7 +1398,7 @@ function preRefSc(toolName, a) {
     reference_videos: vid,
     reference_audio: aud,
     local_references: local,
-    settings: {
+    settings: withSpeech({
       duration: a.duration,
       resolution: a.draft === true ? '480p-draft' : a.draft === false ? String(a.resolution || '').replace(/-draft$/i, '') : a.resolution,
       aspect_ratio: a.aspect_ratio,
@@ -1369,7 +1409,7 @@ function preRefSc(toolName, a) {
       moodboard_ids: a.moodboard_ids,
       moodboard_id: a.moodboard_id,
       preset_id: a.preset_id
-    }
+    }, a)
   };
 }
 
@@ -1427,7 +1467,7 @@ function liveFromTimedOut(sc) {
     kind: kindFromTool(originTool, sc),
     prompt: originArgs.prompt || originArgs.text || '',
     model: sc.model || originArgs.model,
-    settings: {
+    settings: withSpeech({
       duration: originArgs.duration,
       is_draft: originArgs.draft,
       resolution: originArgs.draft === true ? '480p-draft' : originArgs.draft === false ? String(originArgs.resolution || '').replace(/-draft$/i, '') : originArgs.resolution,
@@ -1436,7 +1476,7 @@ function liveFromTimedOut(sc) {
       mode: originArgs.operation,
       visual_dna_ids: originArgs.visual_dna_ids,
       moodboard_id: originArgs.moodboard_id
-    },
+    }, originArgs),
     count: ids ? ids.length : (originArgs.num_images || 1),
     generation_id: sc.generation_id,
     poll_tool: 'get_generation_status',
@@ -1457,14 +1497,14 @@ function completedFromPlain(sc) {
     kind: kindFromTool(originTool, sc),
     prompt: originArgs.prompt || originArgs.text || sc.prompt_used || '',
     model: sc.model || originArgs.model,
-    settings: {
+    settings: withSpeech({
       duration: sc.duration || originArgs.duration,
       is_draft: originArgs.draft,
       resolution: originArgs.draft === true ? '480p-draft' : originArgs.draft === false ? String(originArgs.resolution || '').replace(/-draft$/i, '') : originArgs.resolution,
       aspect_ratio: originArgs.aspect_ratio,
       quality: originArgs.quality,
       mode: originArgs.operation
-    },
+    }, originArgs),
     urls: sc.urls || [],
     session_id: sc.session_id || originArgs.session_id,
     project_id: sc.project_id || originArgs.project_id,

package/src/tools/_shared.js

@@ -1066,6 +1066,9 @@ async function uiCompleted(p, textPayload, extraContent) {
     // 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 } : {}),
+    // Speech settings the API reports it applied (generate.js speechFromResult);
+    // the card prefers them over the submitted `settings`.
+    ...(p.speech ? { speech: p.speech } : {}),
     // 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

@@ -151,6 +151,24 @@ function mediaKind(url) {
   return 'image';
 }
 
+// The speech settings the API says it APPLIED (status result, kolbo-api
+// sdk/controller.js extractResult). Keys are always present (null = not
+// applied) so the card prefers these over what the caller asked for; only a
+// speech result carries `enforce_language`, so every other result returns
+// undefined and its card is untouched.
+function speechFromResult(r) {
+  if (!r || typeof r !== 'object' || !Object.prototype.hasOwnProperty.call(r, 'enforce_language')) return undefined;
+  const vs = r.voice_settings || {};
+  const out = {
+    stability: vs.stability, similarity_boost: vs.similarity_boost, style: vs.style, use_speaker_boost: vs.use_speaker_boost,
+    speaking_speed: r.speaking_speed, enforce_language: r.enforce_language, output_format: r.output_format,
+    text_normalization: r.text_normalization, audio_effect: r.audio_effect,
+    audio_effect_strength: r.audio_effect_strength, seed: r.seed,
+  };
+  for (const k of Object.keys(out)) if (out[k] === undefined) out[k] = null;
+  return out;
+}
+
 function preferOwned(urls) {
   const list = (urls || []).filter((item) => typeof item === 'string' && item);
   const ours = list.filter(ownedUrl);
@@ -1099,14 +1117,20 @@ function registerGenerateTools(server, client, options = {}) {
       emotion: z.string().optional().describe('MiniMax / Cartesia voices. Emotion: happy, sad, angry, fearful, disgusted, surprised, calm, fluent, whisper. Prefer this OR selected_style (both map to the same delivery control).'),
       speaking_speed: z.number().optional().describe('Speech speed 0.5 (slow) – 2.0 (fast); Cartesia clamped 0.6–1.5. Default 1.0. Applies to ElevenLabs / OpenAI / Google / MiniMax / Cartesia / DeepDub (as tempo when tempo omitted).'),
       // ── ElevenLabs voice settings ──
-      similarity_boost: z.number().optional().describe('ElevenLabs voice similarity, 0–1. Default 0.75. Higher hews closer to the original voice.'),
-      style: z.number().optional().describe('ElevenLabs style exaggeration, 0–1. Default 0.5. Higher = more expressive/dramatic.'),
-      use_speaker_boost: z.boolean().optional().describe('ElevenLabs speaker boost. Default true.'),
+      stability: z.number().min(0).max(1).optional().describe('ElevenLabs Stability, 0–1. Default 0.5. Lower = more expressive and varied between takes (Creative); higher = steadier, more consistent delivery (Robust). Supported on V4.'),
+      similarity_boost: z.number().optional().describe('ElevenLabs voice similarity, 0–1. Default 0.75. Higher hews closer to the original voice. Also applies to the user\'s own cloned/character ElevenLabs voices.'),
+      style: z.number().optional().describe('ElevenLabs style exaggeration, 0–1. Default 0. Higher = more expressive/dramatic. V3/V2 voices only — ElevenLabs V4 ignores it (use stability or inline audio tags instead).'),
+      use_speaker_boost: z.boolean().optional().describe('ElevenLabs speaker boost (presence/clarity). Default true.'),
+      enforce_language: z.string().optional().describe('ElevenLabs voices ONLY. Force the spoken language with an ISO 639-1 code ("he", "en", "fr"; a locale like "he-IL" is also accepted). Omit (recommended) and ElevenLabs auto-detects the language from the text. Different from `language`, which stays the accent/locale control.'),
+      output_format: z.enum(['mp3_44100_128', 'mp3_44100_192', 'wav_44100']).optional().describe('ElevenLabs voices ONLY. Audio file format: "mp3_44100_128" (default), "mp3_44100_192" (higher-bitrate MP3) or "wav_44100" (uncompressed; the result is a .wav file). ElevenLabs\' native render is delivered as is. Any other provider rejects it (OUTPUT_FORMAT_UNSUPPORTED).'),
+      text_normalization: z.enum(['auto', 'on', 'off']).optional().describe('ElevenLabs voices. Spell out numbers, dates and abbreviations before speaking: "auto" (default), "on" (always), "off" (read as written).'),
+      audio_effect: z.enum(['none', 'phone', 'radio', 'megaphone', 'small_room', 'hall', 'distant']).optional().describe('Kolbo post-processing effect applied to the finished voice: phone, radio, megaphone, small_room, hall, distant, or "none" (default). Works for ANY voice/provider, no extra credits.'),
+      audio_effect_strength: z.number().min(0).max(1).optional().describe('Mix of audio_effect with the dry voice, 0–1. Default 1 (full effect). Ignored without audio_effect.'),
       // ── DeepDub controls ──
       variance: z.number().optional().describe('DeepDub voice variance, 0–1. Default 0.2. Higher = more takes/variation.'),
       tempo: z.number().optional().describe('DeepDub tempo multiplier, 0–2. Default 1.0.'),
       promptBoost: z.boolean().optional().describe('DeepDub prompt-fidelity boost. Default true.'),
-      seed: z.number().optional().describe('Reproducibility seed (DeepDub / Zonos). Same seed + inputs → same output.'),
+      seed: z.number().optional().describe('Reproducibility seed (ElevenLabs / DeepDub / Zonos). Same seed + inputs → same output. ElevenLabs: integer 0–4294967295, best-effort determinism (not guaranteed).'),
       accentControl: z.object({
         accentBaseLocale: z.string().describe('Base accent locale, e.g. "en-US".'),
         accentLocale: z.string().describe('Target accent locale, e.g. "en-GB".'),
@@ -1121,7 +1145,7 @@ function registerGenerateTools(server, client, options = {}) {
       project_id: projectIdField,
       session_id: sessionIdField
     },
-    async ({ text, voice, model, turbo, language, style_instructions_preset_id, style_instructions, style_instructions_label, selected_style, emotion, speaking_speed, similarity_boost, style, use_speaker_boost, variance, tempo, promptBoost, seed, accentControl, voiceTitle, minimax_pitch, minimax_vol, minimax_intensity, minimax_timbre, project_id, session_id }) => {
+    async ({ text, voice, model, turbo, language, style_instructions_preset_id, style_instructions, style_instructions_label, selected_style, emotion, speaking_speed, stability, similarity_boost, style, use_speaker_boost, enforce_language, output_format, text_normalization, audio_effect, audio_effect_strength, variance, tempo, promptBoost, seed, accentControl, voiceTitle, minimax_pitch, minimax_vol, minimax_intensity, minimax_timbre, project_id, session_id }) => {
       model = await canonicalModelId(client, model, 'text_to_speech'); // lenient id resolution ("z-image" → "z-image/turbo")
       // Resolve the requested voice against the REAL catalog (cached) so the card
       // can show its display name + portrait instead of a raw id, and so an id
@@ -1133,11 +1157,14 @@ function registerGenerateTools(server, client, options = {}) {
       const unknownVoice = voice && !voiceRecord && !/^custom_/i.test(voice)
         ? `Voice "${voice}" is not in the Kolbo voice catalog. Call list_voices and pass a voice_id it returns — an unrecognised id is NOT rejected, it is silently mapped to a different voice, so the audio will not be the voice you named.`
         : null;
+      // ElevenLabs / effect controls carried on the card (undefined keys drop out of the JSON).
+      const ttsControls = { turbo, stability, similarity_boost, style, use_speaker_boost, enforce_language, output_format, text_normalization, audio_effect, audio_effect_strength, seed };
       const gen = await client.post('/v1/generate/speech', {
         text, voice, model, turbo, language,
         style_instructions_preset_id, style_instructions, style_instructions_label,
         selected_style, emotion, speaking_speed,
-        similarity_boost, style, use_speaker_boost,
+        stability, similarity_boost, style, use_speaker_boost,
+        enforce_language, output_format, text_normalization, audio_effect, audio_effect_strength,
         variance, tempo, promptBoost, seed, accentControl, voiceTitle,
         minimax_pitch, minimax_vol, minimax_intensity, minimax_timbre,
         project_id, session_id
@@ -1151,6 +1178,7 @@ function registerGenerateTools(server, client, options = {}) {
           style: selected_style || emotion || style_instructions_preset_id || style_instructions,
           speaking_speed,
           language,
+          ...ttsControls,
         },
         warning: unknownVoice
       });
@@ -1178,7 +1206,9 @@ function registerGenerateTools(server, client, options = {}) {
           style: selected_style || emotion || style_instructions_preset_id || style_instructions,
           speaking_speed,
           language,
+          ...ttsControls,
         },
+        speech: speechFromResult(result.result),
         urls: result.result.urls,
         playback_urls: result.result.playback_urls,
         duration: result.result.duration,
@@ -1420,6 +1450,7 @@ function registerGenerateTools(server, client, options = {}) {
           txt_url: res.txt_url || undefined,
           audio_url: res.audio_url || undefined,
           prompt: res.prompt_used || res.prompt || undefined,
+          speech: speechFromResult(res),
           credits_used: creditFields(single).credits_used,
           items: [{
             id: single.generation_id,