@kolbo/mcp 1.111.0 → 1.112.0

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 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_speech` | Text + voice → spoken audio. `eleven_model` (eleven_v4 / eleven_v3 / eleven_multilingual_v2) picks which ElevenLabs model speaks an ElevenLabs voice (stock or own clone; omit = the voice's default; same price; `turbo` needs v4). 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": "1.111.0",
+  "version": "1.112.0",
   "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

@@ -233,8 +233,9 @@ function voiceLabel(sc) { return sc.voice_name || sc.voice || (sc.settings || {}
 // 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',
+var SPEECH_KEYS = ['eleven_model', 'turbo', 'speaking_speed', 'stability', 'similarity_boost', 'enforce_language',
   'output_format', 'audio_effect', 'audio_effect_strength', 'seed'];
+var SPEECH_MODELS = { eleven_v4: 'V4', eleven_v3: 'V3', eleven_multilingual_v2: 'V2' };
 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]; });
@@ -251,6 +252,7 @@ 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;
+  v = speechVal(sc, 'eleven_model'); if (v) h += chip('Model ' + esc(SPEECH_MODELS[v] || 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)));

package/src/tools/_shared.js

@@ -1066,7 +1066,8 @@ 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);
+    // Speech settings the API reports it applied (generate.js speechFromResult,
+    // incl. eleven_model); passed through whole.
     // 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

package/src/tools/generate.js

@@ -163,7 +163,7 @@ function speechFromResult(r) {
     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,
+    audio_effect_strength: r.audio_effect_strength, seed: r.seed, eleven_model: r.eleven_model,
   };
   for (const k of Object.keys(out)) if (out[k] === undefined) out[k] = null;
   return out;
@@ -1100,8 +1100,9 @@ function registerGenerateTools(server, client, options = {}) {
     {
       text: z.string().describe('The text to convert to speech'),
       voice: z.string().optional().describe('Voice ID or display name — MUST come from a `list_voices` result, never constructed. Google/Gemini ids in particular are not validated provider-side: an id that is not in the catalog is silently mapped to another voice (or a default one) and the audio comes back in a voice nobody asked for. Do not pattern-match a locale onto an id you saw for another language. Default: "Rachel"'),
-      model: z.string().optional().describe('Model identifier. Use list_models type="text_to_speech" to see options. Default: eleven_v4 (ElevenLabs V4). ElevenLabs V4/V3 voices follow inline audio tags in the text: delivery ([whispers], [shouting], [sarcastic]), reactions ([laughs], [sighs], [clears throat]), pauses ([pause], [long pause]) and, on V4, free-form directions like [Low, steady voice, restrained urgency]. Tags are always English, even in Hebrew text. SSML such as <break> is NOT supported.'),
+      model: z.string().optional().describe('Model identifier. Use list_models type="text_to_speech" to see options. Default: eleven_v4 (ElevenLabs V4). To choose which ElevenLabs model speaks a given ElevenLabs voice (stock or a cloned voice), use `eleven_model` instead. ElevenLabs V4/V3 voices follow inline audio tags in the text: delivery ([whispers], [shouting], [sarcastic]), reactions ([laughs], [sighs], [clears throat]), pauses ([pause], [long pause]) and, on V4, free-form directions like [Low, steady voice, restrained urgency]. Tags are always English, even in Hebrew text. SSML such as <break> is NOT supported.'),
       turbo: z.boolean().optional().describe('ElevenLabs V4 voices ONLY. Turbo: faster generation at half the credits, slightly lower quality. Rejected (TURBO_UNSUPPORTED) for any voice not on ElevenLabs V4. Default false.'),
+      eleven_model: z.enum(['eleven_v4', 'eleven_v3', 'eleven_multilingual_v2']).optional().describe("ElevenLabs voices ONLY (stock voices and the user's own cloned/character ElevenLabs voices). Which ElevenLabs model speaks: eleven_v4 (default for new voices, most expressive, audio tags, Stability/Similarity/Turbo), eleven_v3 (audio tags), eleven_multilingual_v2 (no audio tags; supports `style`). Omit = the voice's own default. Same price for all. Rejected with 400 ELEVEN_MODEL_UNSUPPORTED for non-ElevenLabs voices. `turbo` requires eleven_v4."),
       language: z.string().optional().describe('Language / accent code (e.g., "en-US", "he-IL", "es-ES"). For Google/Gemini voices this is the Accent control (does not translate the text). Default: "en-US"'),
       // ── Expressive style / emotion (provider-specific) ──
       // Google/Gemini: prefer a named preset id from the Kolbo UI picker; free-form
@@ -1145,7 +1146,7 @@ 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, 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 }) => {
+    async ({ text, voice, model, eleven_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
@@ -1158,9 +1159,9 @@ function registerGenerateTools(server, client, options = {}) {
         ? `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 ttsControls = { eleven_model, 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,
+        text, voice, model, eleven_model, turbo, language,
         style_instructions_preset_id, style_instructions, style_instructions_label,
         selected_style, emotion, speaking_speed,
         stability, similarity_boost, style, use_speaker_boost,