llmist 2.3.0 → 2.4.0

package/dist/cli.cjs

@@ -2498,7 +2498,27 @@ var init_cost_reporting_client = __esm({
       constructor(client, reportCost) {
         this.client = client;
         this.reportCost = reportCost;
+        this.image = {
+          generate: async (options) => {
+            const result = await this.client.image.generate(options);
+            if (result.cost !== void 0 && result.cost > 0) {
+              this.reportCost(result.cost);
+            }
+            return result;
+          }
+        };
+        this.speech = {
+          generate: async (options) => {
+            const result = await this.client.speech.generate(options);
+            if (result.cost !== void 0 && result.cost > 0) {
+              this.reportCost(result.cost);
+            }
+            return result;
+          }
+        };
       }
+      image;
+      speech;
       /**
        * Access to model registry for cost estimation.
        */
@@ -4591,6 +4611,28 @@ var init_anthropic = __esm({
       getModelSpecs() {
         return ANTHROPIC_MODELS;
       }
+      // =========================================================================
+      // Image Generation (Not Supported)
+      // =========================================================================
+      supportsImageGeneration(_modelId) {
+        return false;
+      }
+      async generateImage() {
+        throw new Error(
+          "Anthropic does not support image generation. Use OpenAI (DALL-E, GPT Image) or Google Gemini (Imagen) instead."
+        );
+      }
+      // =========================================================================
+      // Speech Generation (Not Supported)
+      // =========================================================================
+      supportsSpeechGeneration(_modelId) {
+        return false;
+      }
+      async generateSpeech() {
+        throw new Error(
+          "Anthropic does not support speech generation. Use OpenAI (TTS) or Google Gemini (TTS) instead."
+        );
+      }
       buildRequestPayload(options, descriptor, spec, messages) {
         const systemMessages = messages.filter((message) => message.role === "system");
         const system = systemMessages.length > 0 ? systemMessages.map((m, index) => ({
@@ -4745,6 +4787,182 @@ var init_anthropic = __esm({
   }
 });
 
+// src/providers/gemini-image-models.ts
+function getGeminiImageModelSpec(modelId) {
+  return geminiImageModels.find((m) => m.modelId === modelId);
+}
+function isGeminiImageModel(modelId) {
+  return geminiImageModels.some((m) => m.modelId === modelId);
+}
+function calculateGeminiImageCost(modelId, size = "1:1", n = 1) {
+  const spec = getGeminiImageModelSpec(modelId);
+  if (!spec) return void 0;
+  if (spec.pricing.perImage !== void 0) {
+    return spec.pricing.perImage * n;
+  }
+  if (spec.pricing.bySize) {
+    const sizePrice = spec.pricing.bySize[size];
+    if (typeof sizePrice === "number") {
+      return sizePrice * n;
+    }
+  }
+  return void 0;
+}
+var IMAGEN4_ASPECT_RATIOS, GEMINI_IMAGE_ASPECT_RATIOS, geminiImageModels;
+var init_gemini_image_models = __esm({
+  "src/providers/gemini-image-models.ts"() {
+    "use strict";
+    IMAGEN4_ASPECT_RATIOS = ["1:1", "3:4", "4:3", "9:16", "16:9"];
+    GEMINI_IMAGE_ASPECT_RATIOS = ["1:1", "3:4", "4:3", "9:16", "16:9"];
+    geminiImageModels = [
+      // Imagen 4 Family (standalone image generation)
+      {
+        provider: "gemini",
+        modelId: "imagen-4.0-fast-generate-001",
+        displayName: "Imagen 4 Fast",
+        pricing: {
+          perImage: 0.02
+        },
+        supportedSizes: [...IMAGEN4_ASPECT_RATIOS],
+        maxImages: 4,
+        defaultSize: "1:1",
+        features: {
+          textRendering: true
+        }
+      },
+      {
+        provider: "gemini",
+        modelId: "imagen-4.0-generate-001",
+        displayName: "Imagen 4",
+        pricing: {
+          perImage: 0.04
+        },
+        supportedSizes: [...IMAGEN4_ASPECT_RATIOS],
+        maxImages: 4,
+        defaultSize: "1:1",
+        features: {
+          textRendering: true
+        }
+      },
+      {
+        provider: "gemini",
+        modelId: "imagen-4.0-ultra-generate-001",
+        displayName: "Imagen 4 Ultra",
+        pricing: {
+          perImage: 0.06
+        },
+        supportedSizes: [...IMAGEN4_ASPECT_RATIOS],
+        maxImages: 4,
+        defaultSize: "1:1",
+        features: {
+          textRendering: true
+        }
+      },
+      // Preview versions
+      {
+        provider: "gemini",
+        modelId: "imagen-4.0-generate-preview-06-06",
+        displayName: "Imagen 4 (Preview)",
+        pricing: {
+          perImage: 0.04
+        },
+        supportedSizes: [...IMAGEN4_ASPECT_RATIOS],
+        maxImages: 4,
+        defaultSize: "1:1",
+        features: {
+          textRendering: true
+        }
+      },
+      {
+        provider: "gemini",
+        modelId: "imagen-4.0-ultra-generate-preview-06-06",
+        displayName: "Imagen 4 Ultra (Preview)",
+        pricing: {
+          perImage: 0.06
+        },
+        supportedSizes: [...IMAGEN4_ASPECT_RATIOS],
+        maxImages: 4,
+        defaultSize: "1:1",
+        features: {
+          textRendering: true
+        }
+      },
+      // Gemini Native Image Generation (multimodal models)
+      {
+        provider: "gemini",
+        modelId: "gemini-2.5-flash-image",
+        displayName: "Gemini 2.5 Flash Image",
+        pricing: {
+          perImage: 0.039
+        },
+        supportedSizes: [...GEMINI_IMAGE_ASPECT_RATIOS],
+        maxImages: 1,
+        defaultSize: "1:1",
+        features: {
+          conversational: true,
+          textRendering: true
+        }
+      },
+      {
+        provider: "gemini",
+        modelId: "gemini-2.5-flash-image-preview",
+        displayName: "Gemini 2.5 Flash Image (Preview)",
+        pricing: {
+          perImage: 0.039
+        },
+        supportedSizes: [...GEMINI_IMAGE_ASPECT_RATIOS],
+        maxImages: 1,
+        defaultSize: "1:1",
+        features: {
+          conversational: true,
+          textRendering: true
+        }
+      },
+      {
+        provider: "gemini",
+        modelId: "gemini-3-pro-image-preview",
+        displayName: "Gemini 3 Pro Image (Preview)",
+        pricing: {
+          // Token-based: ~$0.134 per 1K/2K image, $0.24 per 4K
+          // Using 2K as default
+          bySize: {
+            "1K": 0.134,
+            "2K": 0.134,
+            "4K": 0.24
+          }
+        },
+        supportedSizes: ["1K", "2K", "4K"],
+        maxImages: 1,
+        defaultSize: "2K",
+        features: {
+          conversational: true,
+          textRendering: true
+        }
+      },
+      // Alias: nano-banana-pro-preview is gemini-3-pro-image-preview
+      {
+        provider: "gemini",
+        modelId: "nano-banana-pro-preview",
+        displayName: "Nano Banana Pro (Gemini 3 Pro Image)",
+        pricing: {
+          bySize: {
+            "1K": 0.134,
+            "2K": 0.134,
+            "4K": 0.24
+          }
+        },
+        supportedSizes: ["1K", "2K", "4K"],
+        maxImages: 1,
+        defaultSize: "2K",
+        features: {
+          conversational: true,
+          textRendering: true
+        }
+      }
+    ];
+  }
+});
+
 // src/providers/gemini-models.ts
 var GEMINI_MODELS;
 var init_gemini_models = __esm({
@@ -4918,7 +5136,171 @@ var init_gemini_models = __esm({
   }
 });
 
+// src/providers/gemini-speech-models.ts
+function getGeminiSpeechModelSpec(modelId) {
+  return geminiSpeechModels.find((m) => m.modelId === modelId);
+}
+function isGeminiSpeechModel(modelId) {
+  return geminiSpeechModels.some((m) => m.modelId === modelId);
+}
+function calculateGeminiSpeechCost(modelId, characterCount, estimatedMinutes) {
+  const spec = getGeminiSpeechModelSpec(modelId);
+  if (!spec) return void 0;
+  if (spec.pricing.perMinute !== void 0) {
+    if (estimatedMinutes !== void 0) {
+      return estimatedMinutes * spec.pricing.perMinute;
+    }
+    const approxMinutes = characterCount / 750;
+    return approxMinutes * spec.pricing.perMinute;
+  }
+  return void 0;
+}
+var GEMINI_TTS_VOICES, GEMINI_TTS_FORMATS, geminiSpeechModels;
+var init_gemini_speech_models = __esm({
+  "src/providers/gemini-speech-models.ts"() {
+    "use strict";
+    GEMINI_TTS_VOICES = [
+      "Zephyr",
+      // Bright
+      "Puck",
+      // Upbeat
+      "Charon",
+      // Informative
+      "Kore",
+      // Firm
+      "Fenrir",
+      // Excitable
+      "Leda",
+      // Youthful
+      "Orus",
+      // Firm
+      "Aoede",
+      // Breezy
+      "Callirrhoe",
+      // Easy-going
+      "Autonoe",
+      // Bright
+      "Enceladus",
+      // Breathy
+      "Iapetus",
+      // Clear
+      "Umbriel",
+      // Easy-going
+      "Algieba",
+      // Smooth
+      "Despina",
+      // Smooth
+      "Erinome",
+      // Clear
+      "Algenib",
+      // Gravelly
+      "Rasalgethi",
+      // Informative
+      "Laomedeia",
+      // Upbeat
+      "Achernar",
+      // Soft
+      "Alnilam",
+      // Firm
+      "Schedar",
+      // Even
+      "Gacrux",
+      // Mature
+      "Pulcherrima",
+      // Forward
+      "Achird",
+      // Friendly
+      "Zubenelgenubi",
+      // Casual
+      "Vindemiatrix",
+      // Gentle
+      "Sadachbia",
+      // Lively
+      "Sadaltager",
+      // Knowledgeable
+      "Sulafat"
+      // Warm
+    ];
+    GEMINI_TTS_FORMATS = ["pcm", "wav"];
+    geminiSpeechModels = [
+      {
+        provider: "gemini",
+        modelId: "gemini-2.5-flash-preview-tts",
+        displayName: "Gemini 2.5 Flash TTS (Preview)",
+        pricing: {
+          // $0.50 per 1M input tokens = $0.0000005 per token
+          perInputToken: 5e-7,
+          // $10.00 per 1M audio output tokens = $0.00001 per token
+          perAudioOutputToken: 1e-5,
+          // Rough estimate: ~$0.01 per minute of audio
+          perMinute: 0.01
+        },
+        voices: [...GEMINI_TTS_VOICES],
+        formats: GEMINI_TTS_FORMATS,
+        maxInputLength: 8e3,
+        // bytes (text + prompt combined)
+        defaultVoice: "Zephyr",
+        defaultFormat: "wav",
+        features: {
+          multiSpeaker: true,
+          languages: 24,
+          voiceInstructions: true
+        }
+      },
+      {
+        provider: "gemini",
+        modelId: "gemini-2.5-pro-preview-tts",
+        displayName: "Gemini 2.5 Pro TTS (Preview)",
+        pricing: {
+          // $1.00 per 1M input tokens = $0.000001 per token
+          perInputToken: 1e-6,
+          // $20.00 per 1M audio output tokens = $0.00002 per token
+          perAudioOutputToken: 2e-5,
+          // Rough estimate: ~$0.02 per minute of audio
+          perMinute: 0.02
+        },
+        voices: [...GEMINI_TTS_VOICES],
+        formats: GEMINI_TTS_FORMATS,
+        maxInputLength: 8e3,
+        // bytes
+        defaultVoice: "Zephyr",
+        defaultFormat: "wav",
+        features: {
+          multiSpeaker: true,
+          languages: 24,
+          voiceInstructions: true
+        }
+      }
+    ];
+  }
+});
+
 // src/providers/gemini.ts
+function wrapPcmInWav(pcmData, sampleRate, bitsPerSample, numChannels) {
+  const byteRate = sampleRate * numChannels * bitsPerSample / 8;
+  const blockAlign = numChannels * bitsPerSample / 8;
+  const dataSize = pcmData.length;
+  const headerSize = 44;
+  const fileSize = headerSize + dataSize - 8;
+  const buffer = new ArrayBuffer(headerSize + dataSize);
+  const view = new DataView(buffer);
+  const uint8 = new Uint8Array(buffer);
+  view.setUint32(0, 1380533830, false);
+  view.setUint32(4, fileSize, true);
+  view.setUint32(8, 1463899717, false);
+  view.setUint32(12, 1718449184, false);
+  view.setUint32(16, 16, true);
+  view.setUint16(20, 1, true);
+  view.setUint16(22, numChannels, true);
+  view.setUint32(24, sampleRate, true);
+  view.setUint32(28, byteRate, true);
+  view.setUint16(32, blockAlign, true);
+  view.setUint16(34, bitsPerSample, true);
+  view.setUint32(36, 1684108385, false);
+  view.setUint32(40, dataSize, true);
+  uint8.set(pcmData, headerSize);
+  return buffer;
+}
 function createGeminiProviderFromEnv() {
   return createProviderFromEnv("GEMINI_API_KEY", import_genai.GoogleGenAI, GeminiGenerativeProvider);
 }
@@ -4929,7 +5311,9 @@ var init_gemini = __esm({
     import_genai = require("@google/genai");
     init_base_provider();
     init_constants2();
+    init_gemini_image_models();
     init_gemini_models();
+    init_gemini_speech_models();
     init_utils();
     GEMINI_ROLE_MAP = {
       system: "user",
@@ -4944,6 +5328,139 @@ var init_gemini = __esm({
       getModelSpecs() {
         return GEMINI_MODELS;
       }
+      // =========================================================================
+      // Image Generation
+      // =========================================================================
+      getImageModelSpecs() {
+        return geminiImageModels;
+      }
+      supportsImageGeneration(modelId) {
+        return isGeminiImageModel(modelId);
+      }
+      async generateImage(options) {
+        const client = this.client;
+        const spec = getGeminiImageModelSpec(options.model);
+        const isImagenModel = options.model.startsWith("imagen");
+        const aspectRatio = options.size ?? spec?.defaultSize ?? "1:1";
+        const n = options.n ?? 1;
+        if (isImagenModel) {
+          const response2 = await client.models.generateImages({
+            model: options.model,
+            prompt: options.prompt,
+            config: {
+              numberOfImages: n,
+              aspectRatio,
+              outputMimeType: options.responseFormat === "b64_json" ? "image/png" : "image/jpeg"
+            }
+          });
+          const images2 = response2.generatedImages ?? [];
+          const cost2 = calculateGeminiImageCost(options.model, aspectRatio, images2.length);
+          return {
+            // Gemini's imageBytes is already base64 encoded, so use it directly
+            images: images2.map((img) => ({
+              b64Json: img.image?.imageBytes ?? void 0
+            })),
+            model: options.model,
+            usage: {
+              imagesGenerated: images2.length,
+              size: aspectRatio,
+              quality: "standard"
+            },
+            cost: cost2
+          };
+        }
+        const response = await client.models.generateContent({
+          model: options.model,
+          contents: [{ role: "user", parts: [{ text: options.prompt }] }],
+          config: {
+            responseModalities: [import_genai.Modality.IMAGE, import_genai.Modality.TEXT]
+          }
+        });
+        const images = [];
+        const candidate = response.candidates?.[0];
+        if (candidate?.content?.parts) {
+          for (const part of candidate.content.parts) {
+            if ("inlineData" in part && part.inlineData) {
+              images.push({
+                b64Json: part.inlineData.data
+              });
+            }
+          }
+        }
+        const cost = calculateGeminiImageCost(options.model, aspectRatio, images.length);
+        return {
+          images,
+          model: options.model,
+          usage: {
+            imagesGenerated: images.length,
+            size: aspectRatio,
+            quality: "standard"
+          },
+          cost
+        };
+      }
+      // =========================================================================
+      // Speech Generation
+      // =========================================================================
+      getSpeechModelSpecs() {
+        return geminiSpeechModels;
+      }
+      supportsSpeechGeneration(modelId) {
+        return isGeminiSpeechModel(modelId);
+      }
+      async generateSpeech(options) {
+        const client = this.client;
+        const spec = getGeminiSpeechModelSpec(options.model);
+        const voice = options.voice ?? spec?.defaultVoice ?? "Zephyr";
+        const response = await client.models.generateContent({
+          model: options.model,
+          contents: [
+            {
+              role: "user",
+              parts: [{ text: options.input }]
+            }
+          ],
+          config: {
+            responseModalities: [import_genai.Modality.AUDIO],
+            speechConfig: {
+              voiceConfig: {
+                prebuiltVoiceConfig: {
+                  voiceName: voice
+                }
+              }
+            }
+          }
+        });
+        let pcmData;
+        const candidate = response.candidates?.[0];
+        if (candidate?.content?.parts) {
+          for (const part of candidate.content.parts) {
+            if ("inlineData" in part && part.inlineData?.data) {
+              const base64 = part.inlineData.data;
+              const binary = atob(base64);
+              pcmData = new Uint8Array(binary.length);
+              for (let i = 0; i < binary.length; i++) {
+                pcmData[i] = binary.charCodeAt(i);
+              }
+              break;
+            }
+          }
+        }
+        if (!pcmData) {
+          throw new Error("No audio data in Gemini TTS response");
+        }
+        const audioData = wrapPcmInWav(pcmData, 24e3, 16, 1);
+        const cost = calculateGeminiSpeechCost(options.model, options.input.length);
+        return {
+          audio: audioData,
+          model: options.model,
+          usage: {
+            characterCount: options.input.length
+          },
+          cost,
+          format: spec?.defaultFormat ?? "wav"
+        };
+      }
       buildRequestPayload(options, descriptor, _spec, messages) {
         const contents = this.convertMessagesToContents(messages);
         const generationConfig = this.buildGenerationConfig(options);
@@ -5139,6 +5656,121 @@ var init_gemini = __esm({
   }
 });
 
+// src/providers/openai-image-models.ts
+function getOpenAIImageModelSpec(modelId) {
+  return openaiImageModels.find((m) => m.modelId === modelId);
+}
+function isOpenAIImageModel(modelId) {
+  return openaiImageModels.some((m) => m.modelId === modelId);
+}
+function calculateOpenAIImageCost(modelId, size, quality = "standard", n = 1) {
+  const spec = getOpenAIImageModelSpec(modelId);
+  if (!spec) return void 0;
+  const sizePrice = spec.pricing.bySize?.[size];
+  if (sizePrice === void 0) return void 0;
+  let pricePerImage;
+  if (typeof sizePrice === "number") {
+    pricePerImage = sizePrice;
+  } else {
+    pricePerImage = sizePrice[quality];
+    if (pricePerImage === void 0) return void 0;
+  }
+  return pricePerImage * n;
+}
+var GPT_IMAGE_SIZES, GPT_IMAGE_QUALITIES, DALLE3_SIZES, DALLE3_QUALITIES, DALLE2_SIZES, openaiImageModels;
+var init_openai_image_models = __esm({
+  "src/providers/openai-image-models.ts"() {
+    "use strict";
+    GPT_IMAGE_SIZES = ["1024x1024", "1024x1536", "1536x1024"];
+    GPT_IMAGE_QUALITIES = ["low", "medium", "high"];
+    DALLE3_SIZES = ["1024x1024", "1024x1792", "1792x1024"];
+    DALLE3_QUALITIES = ["standard", "hd"];
+    DALLE2_SIZES = ["256x256", "512x512", "1024x1024"];
+    openaiImageModels = [
+      // GPT Image 1 Family (flagship)
+      {
+        provider: "openai",
+        modelId: "gpt-image-1",
+        displayName: "GPT Image 1",
+        pricing: {
+          bySize: {
+            "1024x1024": { low: 0.011, medium: 0.04, high: 0.17 },
+            "1024x1536": { low: 0.016, medium: 0.06, high: 0.25 },
+            "1536x1024": { low: 0.016, medium: 0.06, high: 0.25 }
+          }
+        },
+        supportedSizes: [...GPT_IMAGE_SIZES],
+        supportedQualities: [...GPT_IMAGE_QUALITIES],
+        maxImages: 1,
+        defaultSize: "1024x1024",
+        defaultQuality: "medium",
+        features: {
+          textRendering: true,
+          transparency: true
+        }
+      },
+      {
+        provider: "openai",
+        modelId: "gpt-image-1-mini",
+        displayName: "GPT Image 1 Mini",
+        pricing: {
+          bySize: {
+            "1024x1024": { low: 5e-3, medium: 0.02, high: 0.052 },
+            "1024x1536": { low: 75e-4, medium: 0.03, high: 0.078 },
+            "1536x1024": { low: 75e-4, medium: 0.03, high: 0.078 }
+          }
+        },
+        supportedSizes: [...GPT_IMAGE_SIZES],
+        supportedQualities: [...GPT_IMAGE_QUALITIES],
+        maxImages: 1,
+        defaultSize: "1024x1024",
+        defaultQuality: "medium",
+        features: {
+          textRendering: true,
+          transparency: true
+        }
+      },
+      // DALL-E Family
+      {
+        provider: "openai",
+        modelId: "dall-e-3",
+        displayName: "DALL-E 3",
+        pricing: {
+          bySize: {
+            "1024x1024": { standard: 0.04, hd: 0.08 },
+            "1024x1792": { standard: 0.08, hd: 0.12 },
+            "1792x1024": { standard: 0.08, hd: 0.12 }
+          }
+        },
+        supportedSizes: [...DALLE3_SIZES],
+        supportedQualities: [...DALLE3_QUALITIES],
+        maxImages: 1,
+        // DALL-E 3 only supports n=1
+        defaultSize: "1024x1024",
+        defaultQuality: "standard",
+        features: {
+          textRendering: true
+        }
+      },
+      {
+        provider: "openai",
+        modelId: "dall-e-2",
+        displayName: "DALL-E 2 (Legacy)",
+        pricing: {
+          bySize: {
+            "256x256": 0.016,
+            "512x512": 0.018,
+            "1024x1024": 0.02
+          }
+        },
+        supportedSizes: [...DALLE2_SIZES],
+        maxImages: 10,
+        defaultSize: "1024x1024"
+      }
+    ];
+  }
+});
+
 // src/providers/openai-models.ts
 var OPENAI_MODELS;
 var init_openai_models = __esm({
@@ -5503,6 +6135,144 @@ var init_openai_models = __esm({
   }
 });
 
+// src/providers/openai-speech-models.ts
+function getOpenAISpeechModelSpec(modelId) {
+  return openaiSpeechModels.find((m) => m.modelId === modelId);
+}
+function isOpenAISpeechModel(modelId) {
+  return openaiSpeechModels.some((m) => m.modelId === modelId);
+}
+function calculateOpenAISpeechCost(modelId, characterCount, estimatedMinutes) {
+  const spec = getOpenAISpeechModelSpec(modelId);
+  if (!spec) return void 0;
+  if (spec.pricing.perCharacter !== void 0) {
+    return characterCount * spec.pricing.perCharacter;
+  }
+  if (spec.pricing.perMinute !== void 0 && estimatedMinutes !== void 0) {
+    return estimatedMinutes * spec.pricing.perMinute;
+  }
+  if (spec.pricing.perMinute !== void 0) {
+    const approxMinutes = characterCount / 750;
+    return approxMinutes * spec.pricing.perMinute;
+  }
+  return void 0;
+}
+var OPENAI_TTS_VOICES, OPENAI_TTS_EXTENDED_VOICES, OPENAI_TTS_FORMATS, openaiSpeechModels;
+var init_openai_speech_models = __esm({
+  "src/providers/openai-speech-models.ts"() {
+    "use strict";
+    OPENAI_TTS_VOICES = [
+      "alloy",
+      "echo",
+      "fable",
+      "onyx",
+      "nova",
+      "shimmer"
+    ];
+    OPENAI_TTS_EXTENDED_VOICES = [
+      ...OPENAI_TTS_VOICES,
+      "ash",
+      "ballad",
+      "coral",
+      "sage",
+      "verse"
+    ];
+    OPENAI_TTS_FORMATS = ["mp3", "opus", "aac", "flac", "wav", "pcm"];
+    openaiSpeechModels = [
+      // Standard TTS models (character-based pricing)
+      {
+        provider: "openai",
+        modelId: "tts-1",
+        displayName: "TTS-1",
+        pricing: {
+          // $15 per 1M characters = $0.000015 per character
+          perCharacter: 15e-6
+        },
+        voices: [...OPENAI_TTS_VOICES],
+        formats: OPENAI_TTS_FORMATS,
+        maxInputLength: 4096,
+        defaultVoice: "alloy",
+        defaultFormat: "mp3",
+        features: {
+          voiceInstructions: false
+        }
+      },
+      {
+        provider: "openai",
+        modelId: "tts-1-1106",
+        displayName: "TTS-1 (Nov 2023)",
+        pricing: {
+          perCharacter: 15e-6
+        },
+        voices: [...OPENAI_TTS_VOICES],
+        formats: OPENAI_TTS_FORMATS,
+        maxInputLength: 4096,
+        defaultVoice: "alloy",
+        defaultFormat: "mp3",
+        features: {
+          voiceInstructions: false
+        }
+      },
+      {
+        provider: "openai",
+        modelId: "tts-1-hd",
+        displayName: "TTS-1 HD",
+        pricing: {
+          // $30 per 1M characters = $0.00003 per character
+          perCharacter: 3e-5
+        },
+        voices: [...OPENAI_TTS_VOICES],
+        formats: OPENAI_TTS_FORMATS,
+        maxInputLength: 4096,
+        defaultVoice: "alloy",
+        defaultFormat: "mp3",
+        features: {
+          voiceInstructions: false
+        }
+      },
+      {
+        provider: "openai",
+        modelId: "tts-1-hd-1106",
+        displayName: "TTS-1 HD (Nov 2023)",
+        pricing: {
+          perCharacter: 3e-5
+        },
+        voices: [...OPENAI_TTS_VOICES],
+        formats: OPENAI_TTS_FORMATS,
+        maxInputLength: 4096,
+        defaultVoice: "alloy",
+        defaultFormat: "mp3",
+        features: {
+          voiceInstructions: false
+        }
+      },
+      // Token-based TTS model with voice instructions support
+      {
+        provider: "openai",
+        modelId: "gpt-4o-mini-tts",
+        displayName: "GPT-4o Mini TTS",
+        pricing: {
+          // $0.60 per 1M input tokens = $0.0000006 per token
+          perInputToken: 6e-7,
+          // $12 per 1M audio output tokens = $0.000012 per token
+          perAudioOutputToken: 12e-6,
+          // ~$0.015 per minute of audio
+          perMinute: 0.015
+        },
+        voices: [...OPENAI_TTS_EXTENDED_VOICES],
+        formats: OPENAI_TTS_FORMATS,
+        maxInputLength: 2e3,
+        // tokens, not characters
+        defaultVoice: "alloy",
+        defaultFormat: "mp3",
+        features: {
+          voiceInstructions: true
+        }
+      }
+    ];
+  }
+});
+
 // src/providers/openai.ts
 function sanitizeExtra(extra, allowTemperature) {
   if (!extra) {
@@ -5524,7 +6294,9 @@ var init_openai = __esm({
     import_tiktoken = require("tiktoken");
     init_base_provider();
     init_constants2();
+    init_openai_image_models();
     init_openai_models();
+    init_openai_speech_models();
     init_utils();
     ROLE_MAP = {
       system: "system",
@@ -5539,6 +6311,87 @@ var init_openai = __esm({
       getModelSpecs() {
         return OPENAI_MODELS;
       }
+      // =========================================================================
+      // Image Generation
+      // =========================================================================
+      getImageModelSpecs() {
+        return openaiImageModels;
+      }
+      supportsImageGeneration(modelId) {
+        return isOpenAIImageModel(modelId);
+      }
+      async generateImage(options) {
+        const client = this.client;
+        const spec = getOpenAIImageModelSpec(options.model);
+        const size = options.size ?? spec?.defaultSize ?? "1024x1024";
+        const quality = options.quality ?? spec?.defaultQuality ?? "standard";
+        const n = options.n ?? 1;
+        const isDallE2 = options.model === "dall-e-2";
+        const isGptImage = options.model.startsWith("gpt-image");
+        const requestParams = {
+          model: options.model,
+          prompt: options.prompt,
+          size,
+          n
+        };
+        if (!isDallE2 && !isGptImage) {
+          requestParams.quality = quality;
+        }
+        if (isGptImage) {
+        } else if (!isDallE2) {
+          requestParams.response_format = options.responseFormat ?? "url";
+        }
+        const response = await client.images.generate(requestParams);
+        const cost = calculateOpenAIImageCost(options.model, size, quality, n);
+        const images = response.data ?? [];
+        return {
+          images: images.map((img) => ({
+            url: img.url,
+            b64Json: img.b64_json,
+            revisedPrompt: img.revised_prompt
+          })),
+          model: options.model,
+          usage: {
+            imagesGenerated: images.length,
+            size,
+            quality
+          },
+          cost
+        };
+      }
+      // =========================================================================
+      // Speech Generation
+      // =========================================================================
+      getSpeechModelSpecs() {
+        return openaiSpeechModels;
+      }
+      supportsSpeechGeneration(modelId) {
+        return isOpenAISpeechModel(modelId);
+      }
+      async generateSpeech(options) {
+        const client = this.client;
+        const spec = getOpenAISpeechModelSpec(options.model);
+        const format = options.responseFormat ?? spec?.defaultFormat ?? "mp3";
+        const voice = options.voice ?? spec?.defaultVoice ?? "alloy";
+        const response = await client.audio.speech.create({
+          model: options.model,
+          input: options.input,
+          voice,
+          response_format: format,
+          speed: options.speed ?? 1
+        });
+        const audioBuffer = await response.arrayBuffer();
+        const cost = calculateOpenAISpeechCost(options.model, options.input.length);
+        return {
+          audio: audioBuffer,
+          model: options.model,
+          usage: {
+            characterCount: options.input.length
+          },
+          cost,
+          format
+        };
+      }
       buildRequestPayload(options, descriptor, spec, messages) {
         const { maxTokens, temperature, topP, stopSequences, extra } = options;
         const supportsTemperature = spec?.metadata?.supportsTemperature !== false;
@@ -5879,30 +6732,109 @@ var init_model_registry = __esm({
   }
 });
 
-// src/core/options.ts
-var ModelIdentifierParser;
-var init_options = __esm({
-  "src/core/options.ts"() {
+// src/core/namespaces/image.ts
+var ImageNamespace;
+var init_image = __esm({
+  "src/core/namespaces/image.ts"() {
     "use strict";
-    ModelIdentifierParser = class {
-      constructor(defaultProvider = "openai") {
+    ImageNamespace = class {
+      constructor(adapters, defaultProvider) {
+        this.adapters = adapters;
+        this.defaultProvider = defaultProvider;
+      }
+      /**
+       * Generate images from a text prompt.
+       *
+       * @param options - Image generation options
+       * @returns Promise resolving to the generation result with images and cost
+       * @throws Error if the provider doesn't support image generation
+       */
+      async generate(options) {
+        const modelId = options.model;
+        const adapter = this.findImageAdapter(modelId);
+        if (!adapter || !adapter.generateImage) {
+          throw new Error(
+            `No provider supports image generation for model "${modelId}". Available image models: ${this.listModels().map((m) => m.modelId).join(", ")}`
+          );
+        }
+        return adapter.generateImage(options);
+      }
+      /**
+       * List all available image generation models.
+       */
+      listModels() {
+        const models = [];
+        for (const adapter of this.adapters) {
+          if (adapter.getImageModelSpecs) {
+            models.push(...adapter.getImageModelSpecs());
+          }
+        }
+        return models;
+      }
+      /**
+       * Check if a model is supported for image generation.
+       */
+      supportsModel(modelId) {
+        return this.findImageAdapter(modelId) !== void 0;
+      }
+      findImageAdapter(modelId) {
+        return this.adapters.find(
+          (adapter) => adapter.supportsImageGeneration?.(modelId) ?? false
+        );
+      }
+    };
+  }
+});
+
+// src/core/namespaces/speech.ts
+var SpeechNamespace;
+var init_speech = __esm({
+  "src/core/namespaces/speech.ts"() {
+    "use strict";
+    SpeechNamespace = class {
+      constructor(adapters, defaultProvider) {
+        this.adapters = adapters;
         this.defaultProvider = defaultProvider;
       }
-      parse(identifier) {
-        const trimmed = identifier.trim();
-        if (!trimmed) {
-          throw new Error("Model identifier cannot be empty");
-        }
-        const [maybeProvider, ...rest] = trimmed.split(":");
-        if (rest.length === 0) {
-          return { provider: this.defaultProvider, name: maybeProvider };
+      /**
+       * Generate speech audio from text.
+       *
+       * @param options - Speech generation options
+       * @returns Promise resolving to the generation result with audio and cost
+       * @throws Error if the provider doesn't support speech generation
+       */
+      async generate(options) {
+        const modelId = options.model;
+        const adapter = this.findSpeechAdapter(modelId);
+        if (!adapter || !adapter.generateSpeech) {
+          throw new Error(
+            `No provider supports speech generation for model "${modelId}". Available speech models: ${this.listModels().map((m) => m.modelId).join(", ")}`
+          );
         }
-        const provider = maybeProvider;
-        const name = rest.join(":");
-        if (!name) {
-          throw new Error("Model name cannot be empty");
+        return adapter.generateSpeech(options);
+      }
+      /**
+       * List all available speech generation models.
+       */
+      listModels() {
+        const models = [];
+        for (const adapter of this.adapters) {
+          if (adapter.getSpeechModelSpecs) {
+            models.push(...adapter.getSpeechModelSpecs());
+          }
         }
-        return { provider, name };
+        return models;
+      }
+      /**
+       * Check if a model is supported for speech generation.
+       */
+      supportsModel(modelId) {
+        return this.findSpeechAdapter(modelId) !== void 0;
+      }
+      findSpeechAdapter(modelId) {
+        return this.adapters.find(
+          (adapter) => adapter.supportsSpeechGeneration?.(modelId) ?? false
+        );
       }
     };
   }
@@ -5951,6 +6883,69 @@ var init_quick_methods = __esm({
   }
 });
 
+// src/core/namespaces/text.ts
+var TextNamespace;
+var init_text = __esm({
+  "src/core/namespaces/text.ts"() {
+    "use strict";
+    init_quick_methods();
+    TextNamespace = class {
+      constructor(client) {
+        this.client = client;
+      }
+      /**
+       * Generate a complete text response.
+       *
+       * @param prompt - User prompt
+       * @param options - Optional configuration
+       * @returns Complete text response
+       */
+      async complete(prompt, options) {
+        return complete(this.client, prompt, options);
+      }
+      /**
+       * Stream text chunks.
+       *
+       * @param prompt - User prompt
+       * @param options - Optional configuration
+       * @returns Async generator yielding text chunks
+       */
+      stream(prompt, options) {
+        return stream(this.client, prompt, options);
+      }
+    };
+  }
+});
+
+// src/core/options.ts
+var ModelIdentifierParser;
+var init_options = __esm({
+  "src/core/options.ts"() {
+    "use strict";
+    ModelIdentifierParser = class {
+      constructor(defaultProvider = "openai") {
+        this.defaultProvider = defaultProvider;
+      }
+      parse(identifier) {
+        const trimmed = identifier.trim();
+        if (!trimmed) {
+          throw new Error("Model identifier cannot be empty");
+        }
+        const [maybeProvider, ...rest] = trimmed.split(":");
+        if (rest.length === 0) {
+          return { provider: this.defaultProvider, name: maybeProvider };
+        }
+        const provider = maybeProvider;
+        const name = rest.join(":");
+        if (!name) {
+          throw new Error("Model name cannot be empty");
+        }
+        return { provider, name };
+      }
+    };
+  }
+});
+
 // src/core/client.ts
 var client_exports = {};
 __export(client_exports, {
@@ -5963,12 +6958,20 @@ var init_client = __esm({
     init_builder();
     init_discovery();
     init_model_registry();
+    init_image();
+    init_speech();
+    init_text();
     init_options();
     init_quick_methods();
     LLMist = class _LLMist {
       parser;
+      defaultProvider;
       modelRegistry;
       adapters;
+      // Namespaces for different generation types
+      text;
+      image;
+      speech;
       constructor(...args) {
         let adapters = [];
         let defaultProvider;
@@ -6007,6 +7010,7 @@ var init_client = __esm({
           const priorityB = b.priority ?? 0;
           return priorityB - priorityA;
         });
+        this.defaultProvider = resolvedDefaultProvider;
         this.parser = new ModelIdentifierParser(resolvedDefaultProvider);
         this.modelRegistry = new ModelRegistry();
         for (const adapter of this.adapters) {
@@ -6015,6 +7019,9 @@ var init_client = __esm({
         if (customModels.length > 0) {
           this.modelRegistry.registerModels(customModels);
         }
+        this.text = new TextNamespace(this);
+        this.image = new ImageNamespace(this.adapters, this.defaultProvider);
+        this.speech = new SpeechNamespace(this.adapters, this.defaultProvider);
       }
       stream(options) {
         const descriptor = this.parser.parse(options.model);
@@ -6995,7 +8002,9 @@ var COMMANDS = {
   complete: "complete",
   agent: "agent",
   models: "models",
-  gadget: "gadget"
+  gadget: "gadget",
+  image: "image",
+  speech: "speech"
 };
 var LOG_LEVELS = ["silly", "trace", "debug", "info", "warn", "error", "fatal"];
 var DEFAULT_MODEL = "openai:gpt-5-nano";
@@ -7016,7 +8025,17 @@ var OPTION_FLAGS = {
   docker: "--docker",
   dockerRo: "--docker-ro",
   noDocker: "--no-docker",
-  dockerDev: "--docker-dev"
+  dockerDev: "--docker-dev",
+  // Image generation options
+  imageSize: "--size <size>",
+  imageQuality: "--quality <quality>",
+  imageCount: "-n, --count <number>",
+  imageOutput: "-o, --output <path>",
+  // Speech generation options
+  voice: "--voice <name>",
+  speechFormat: "--format <format>",
+  speechSpeed: "--speed <value>",
+  speechOutput: "-o, --output <path>"
 };
 var OPTION_DESCRIPTIONS = {
   model: "Model identifier, e.g. openai:gpt-5-nano or anthropic:claude-sonnet-4-5.",
@@ -7035,7 +8054,17 @@ var OPTION_DESCRIPTIONS = {
   docker: "Run agent in a Docker sandbox container for security isolation.",
   dockerRo: "Run in Docker with current directory mounted read-only.",
   noDocker: "Disable Docker sandboxing (override config).",
-  dockerDev: "Run in Docker dev mode (mount local source instead of npm install)."
+  dockerDev: "Run in Docker dev mode (mount local source instead of npm install).",
+  // Image generation descriptions
+  imageSize: "Image size/aspect ratio, e.g. '1024x1024', '1:1', '16:9'.",
+  imageQuality: "Image quality: 'standard', 'hd', 'low', 'medium', 'high'.",
+  imageCount: "Number of images to generate (model dependent, usually 1-4).",
+  imageOutput: "Output path for the generated image. Defaults to stdout if not specified.",
+  // Speech generation descriptions
+  voice: "Voice name for speech generation, e.g. 'nova', 'alloy', 'Zephyr'.",
+  speechFormat: "Audio format: 'mp3', 'opus', 'aac', 'flac', 'wav', 'pcm'.",
+  speechSpeed: "Speech speed multiplier (0.25 to 4.0, default 1.0).",
+  speechOutput: "Output path for audio file. Defaults to stdout if not specified."
 };
 var SUMMARY_PREFIX = "[llmist]";
 
@@ -7045,7 +8074,7 @@ var import_commander2 = require("commander");
 // package.json
 var package_default = {
   name: "llmist",
-  version: "2.2.0",
+  version: "2.4.0",
   description: "TypeScript LLM client with streaming tool execution. Tools fire mid-stream. Built-in function calling works with any model\u2014no structured outputs or native tool support required.",
   type: "module",
   main: "dist/index.cjs",
@@ -9127,6 +10156,22 @@ var AGENT_CONFIG_KEYS = /* @__PURE__ */ new Set([
   "docker-cwd-permission"
   // Override CWD mount permission for this profile
 ]);
+var IMAGE_CONFIG_KEYS = /* @__PURE__ */ new Set([
+  "model",
+  "size",
+  "quality",
+  "count",
+  "output",
+  "quiet"
+]);
+var SPEECH_CONFIG_KEYS = /* @__PURE__ */ new Set([
+  "model",
+  "voice",
+  "format",
+  "speed",
+  "output",
+  "quiet"
+]);
 var CUSTOM_CONFIG_KEYS = /* @__PURE__ */ new Set([
   ...COMPLETE_CONFIG_KEYS,
   ...AGENT_CONFIG_KEYS,
@@ -9387,6 +10432,75 @@ function validateAgentConfig(raw, section) {
   }
   return result;
 }
+function validateImageConfig(raw, section) {
+  if (typeof raw !== "object" || raw === null) {
+    throw new ConfigError(`[${section}] must be a table`);
+  }
+  const rawObj = raw;
+  for (const key of Object.keys(rawObj)) {
+    if (!IMAGE_CONFIG_KEYS.has(key)) {
+      throw new ConfigError(`[${section}].${key} is not a valid option`);
+    }
+  }
+  const result = {};
+  if ("model" in rawObj) {
+    result.model = validateString(rawObj.model, "model", section);
+  }
+  if ("size" in rawObj) {
+    result.size = validateString(rawObj.size, "size", section);
+  }
+  if ("quality" in rawObj) {
+    result.quality = validateString(rawObj.quality, "quality", section);
+  }
+  if ("count" in rawObj) {
+    result.count = validateNumber(rawObj.count, "count", section, {
+      integer: true,
+      min: 1,
+      max: 10
+    });
+  }
+  if ("output" in rawObj) {
+    result.output = validateString(rawObj.output, "output", section);
+  }
+  if ("quiet" in rawObj) {
+    result.quiet = validateBoolean(rawObj.quiet, "quiet", section);
+  }
+  return result;
+}
+function validateSpeechConfig(raw, section) {
+  if (typeof raw !== "object" || raw === null) {
+    throw new ConfigError(`[${section}] must be a table`);
+  }
+  const rawObj = raw;
+  for (const key of Object.keys(rawObj)) {
+    if (!SPEECH_CONFIG_KEYS.has(key)) {
+      throw new ConfigError(`[${section}].${key} is not a valid option`);
+    }
+  }
+  const result = {};
+  if ("model" in rawObj) {
+    result.model = validateString(rawObj.model, "model", section);
+  }
+  if ("voice" in rawObj) {
+    result.voice = validateString(rawObj.voice, "voice", section);
+  }
+  if ("format" in rawObj) {
+    result.format = validateString(rawObj.format, "format", section);
+  }
+  if ("speed" in rawObj) {
+    result.speed = validateNumber(rawObj.speed, "speed", section, {
+      min: 0.25,
+      max: 4
+    });
+  }
+  if ("output" in rawObj) {
+    result.output = validateString(rawObj.output, "output", section);
+  }
+  if ("quiet" in rawObj) {
+    result.quiet = validateBoolean(rawObj.quiet, "quiet", section);
+  }
+  return result;
+}
 function validateStringOrBoolean(value, field, section) {
   if (typeof value === "string" || typeof value === "boolean") {
     return value;
@@ -9509,6 +10623,10 @@ function validateConfig(raw, configPath) {
         result.complete = validateCompleteConfig(value, key);
       } else if (key === "agent") {
         result.agent = validateAgentConfig(value, key);
+      } else if (key === "image") {
+        result.image = validateImageConfig(value, key);
+      } else if (key === "speech") {
+        result.speech = validateSpeechConfig(value, key);
       } else if (key === "prompts") {
         result.prompts = validatePromptsConfig(value, key);
       } else if (key === "docker") {
@@ -9553,7 +10671,7 @@ function loadConfig() {
   return resolveTemplatesInConfig(inherited, configPath);
 }
 function getCustomCommandNames(config) {
-  const reserved = /* @__PURE__ */ new Set(["global", "complete", "agent", "prompts", "docker"]);
+  const reserved = /* @__PURE__ */ new Set(["global", "complete", "agent", "image", "speech", "prompts", "docker"]);
   return Object.keys(config).filter((key) => !reserved.has(key));
 }
 function resolveTemplatesInConfig(config, configPath) {
@@ -11148,19 +12266,118 @@ function registerGadgetCommand(program, env) {
   );
 }
 
+// src/cli/image-command.ts
+var import_node_fs11 = require("fs");
+var DEFAULT_IMAGE_MODEL = "dall-e-3";
+async function executeImage(promptArg, options, env) {
+  const prompt = await resolvePrompt(promptArg, env);
+  const client = env.createClient();
+  const model = options.model;
+  const n = options.count ? Number.parseInt(options.count, 10) : 1;
+  const stderrTTY = env.stderr.isTTY === true;
+  if (!options.quiet && stderrTTY) {
+    env.stderr.write(`${SUMMARY_PREFIX} Generating image with ${model}...
+`);
+  }
+  const result = await client.image.generate({
+    model,
+    prompt,
+    size: options.size,
+    quality: options.quality,
+    n,
+    responseFormat: options.output ? "b64_json" : "url"
+  });
+  if (options.output) {
+    const imageData = result.images[0];
+    if (imageData.b64Json) {
+      const buffer = Buffer.from(imageData.b64Json, "base64");
+      (0, import_node_fs11.writeFileSync)(options.output, buffer);
+      if (!options.quiet) {
+        env.stderr.write(`${SUMMARY_PREFIX} Image saved to ${options.output}
+`);
+      }
+    } else if (imageData.url) {
+      env.stdout.write(`${imageData.url}
+`);
+    }
+  } else {
+    for (const image of result.images) {
+      if (image.url) {
+        env.stdout.write(`${image.url}
+`);
+      } else if (image.b64Json) {
+        env.stdout.write(image.b64Json);
+      }
+    }
+  }
+  if (!options.quiet && stderrTTY) {
+    const parts = [
+      `${result.images.length} image(s)`,
+      `size: ${result.usage.size}`,
+      `quality: ${result.usage.quality}`
+    ];
+    if (result.cost !== void 0) {
+      parts.push(`cost: ${formatCost(result.cost)}`);
+    }
+    env.stderr.write(`${SUMMARY_PREFIX} ${parts.join(" | ")}
+`);
+  }
+}
+function registerImageCommand(program, env, config) {
+  program.command(COMMANDS.image).description("Generate images from a text prompt.").argument("[prompt]", "Image generation prompt. If omitted, stdin is used when available.").option(
+    OPTION_FLAGS.model,
+    OPTION_DESCRIPTIONS.model,
+    config?.model ?? DEFAULT_IMAGE_MODEL
+  ).option(OPTION_FLAGS.imageSize, OPTION_DESCRIPTIONS.imageSize, config?.size).option(OPTION_FLAGS.imageQuality, OPTION_DESCRIPTIONS.imageQuality, config?.quality).option(OPTION_FLAGS.imageCount, OPTION_DESCRIPTIONS.imageCount, config?.count?.toString()).option(OPTION_FLAGS.imageOutput, OPTION_DESCRIPTIONS.imageOutput, config?.output).option(OPTION_FLAGS.quiet, OPTION_DESCRIPTIONS.quiet, config?.quiet ?? false).action(
+    (prompt, options) => executeAction(() => executeImage(prompt, options, env), env)
+  );
+}
+
 // src/cli/models-command.ts
 var import_chalk8 = __toESM(require("chalk"), 1);
 init_model_shortcuts();
 async function handleModelsCommand(options, env) {
   const client = env.createClient();
-  const models = client.modelRegistry.listModels(options.provider);
+  const showText = options.all || options.text || !options.image && !options.speech;
+  const showImage = options.all || options.image;
+  const showSpeech = options.all || options.speech;
+  const textModels = showText ? client.modelRegistry.listModels(options.provider) : [];
+  const imageModels = showImage ? client.image.listModels().filter((m) => !options.provider || m.provider === options.provider) : [];
+  const speechModels = showSpeech ? client.speech.listModels().filter((m) => !options.provider || m.provider === options.provider) : [];
   if (options.format === "json") {
-    renderJSON(models, env.stdout);
+    renderJSON(textModels, imageModels, speechModels, env.stdout);
   } else {
-    renderTable(models, options.verbose || false, env.stdout);
+    renderAllTables(textModels, imageModels, speechModels, options.verbose || false, env.stdout);
+  }
+}
+function renderAllTables(textModels, imageModels, speechModels, verbose, stream2) {
+  const hasAnyModels = textModels.length > 0 || imageModels.length > 0 || speechModels.length > 0;
+  if (!hasAnyModels) {
+    stream2.write(import_chalk8.default.yellow("\nNo models found matching the specified criteria.\n\n"));
+    return;
+  }
+  stream2.write(import_chalk8.default.bold.cyan("\nAvailable Models\n"));
+  stream2.write(import_chalk8.default.cyan("=".repeat(80)) + "\n\n");
+  if (textModels.length > 0) {
+    renderTextTable(textModels, verbose, stream2);
+  }
+  if (imageModels.length > 0) {
+    renderImageTable(imageModels, verbose, stream2);
+  }
+  if (speechModels.length > 0) {
+    renderSpeechTable(speechModels, verbose, stream2);
+  }
+  if (textModels.length > 0) {
+    stream2.write(import_chalk8.default.bold.magenta("Model Shortcuts\n"));
+    stream2.write(import_chalk8.default.dim("\u2500".repeat(80)) + "\n");
+    const shortcuts = Object.entries(MODEL_ALIASES).sort((a, b) => a[0].localeCompare(b[0]));
+    for (const [shortcut, fullName] of shortcuts) {
+      stream2.write(import_chalk8.default.cyan(`  ${shortcut.padEnd(15)}`) + import_chalk8.default.dim(" \u2192 ") + import_chalk8.default.white(fullName) + "\n");
+    }
+    stream2.write("\n");
   }
 }
-function renderTable(models, verbose, stream2) {
+function renderTextTable(models, verbose, stream2) {
   const grouped = /* @__PURE__ */ new Map();
   for (const model of models) {
     const provider = model.provider;
@@ -11169,13 +12386,13 @@ function renderTable(models, verbose, stream2) {
     }
     grouped.get(provider).push(model);
   }
-  stream2.write(import_chalk8.default.bold.cyan("\nAvailable Models\n"));
-  stream2.write(import_chalk8.default.cyan("=".repeat(80)) + "\n\n");
+  stream2.write(import_chalk8.default.bold.blue("\u{1F4DD} Text/LLM Models\n"));
+  stream2.write(import_chalk8.default.dim("\u2500".repeat(80)) + "\n\n");
   const providers = Array.from(grouped.keys()).sort();
   for (const provider of providers) {
     const providerModels = grouped.get(provider);
     const providerName = provider.charAt(0).toUpperCase() + provider.slice(1);
-    stream2.write(import_chalk8.default.bold.yellow(`${providerName} Models
+    stream2.write(import_chalk8.default.bold.yellow(`${providerName}
 `));
     if (verbose) {
       renderVerboseTable(providerModels, stream2);
@@ -11184,13 +12401,6 @@ function renderTable(models, verbose, stream2) {
     }
     stream2.write("\n");
   }
-  stream2.write(import_chalk8.default.bold.magenta("Model Shortcuts\n"));
-  stream2.write(import_chalk8.default.dim("\u2500".repeat(80)) + "\n");
-  const shortcuts = Object.entries(MODEL_ALIASES).sort((a, b) => a[0].localeCompare(b[0]));
-  for (const [shortcut, fullName] of shortcuts) {
-    stream2.write(import_chalk8.default.cyan(`  ${shortcut.padEnd(15)}`) + import_chalk8.default.dim(" \u2192 ") + import_chalk8.default.white(fullName) + "\n");
-  }
-  stream2.write("\n");
 }
 function renderCompactTable(models, stream2) {
   const idWidth = 25;
@@ -11267,9 +12477,171 @@ function renderVerboseTable(models, stream2) {
   }
   stream2.write("\n");
 }
-function renderJSON(models, stream2) {
-  const output = {
-    models: models.map((model) => ({
+function renderImageTable(models, verbose, stream2) {
+  stream2.write(import_chalk8.default.bold.green("\u{1F3A8} Image Generation Models\n"));
+  stream2.write(import_chalk8.default.dim("\u2500".repeat(80)) + "\n\n");
+  const grouped = /* @__PURE__ */ new Map();
+  for (const model of models) {
+    if (!grouped.has(model.provider)) {
+      grouped.set(model.provider, []);
+    }
+    grouped.get(model.provider).push(model);
+  }
+  for (const [provider, providerModels] of Array.from(grouped.entries()).sort()) {
+    const providerName = provider.charAt(0).toUpperCase() + provider.slice(1);
+    stream2.write(import_chalk8.default.bold.yellow(`${providerName}
+`));
+    if (verbose) {
+      for (const model of providerModels) {
+        stream2.write(import_chalk8.default.bold.green(`
+  ${model.modelId}
+`));
+        stream2.write(import_chalk8.default.dim("  " + "\u2500".repeat(60)) + "\n");
+        stream2.write(`  ${import_chalk8.default.dim("Name:")}      ${import_chalk8.default.white(model.displayName)}
+`);
+        stream2.write(`  ${import_chalk8.default.dim("Sizes:")}     ${import_chalk8.default.yellow(model.supportedSizes.join(", "))}
+`);
+        if (model.supportedQualities) {
+          stream2.write(`  ${import_chalk8.default.dim("Qualities:")} ${import_chalk8.default.yellow(model.supportedQualities.join(", "))}
+`);
+        }
+        stream2.write(`  ${import_chalk8.default.dim("Max Images:")} ${import_chalk8.default.yellow(model.maxImages.toString())}
+`);
+        stream2.write(`  ${import_chalk8.default.dim("Pricing:")}   ${import_chalk8.default.cyan(formatImagePrice(model))}
+`);
+        if (model.features) {
+          const features = [];
+          if (model.features.textRendering) features.push("text-rendering");
+          if (model.features.transparency) features.push("transparency");
+          if (model.features.conversational) features.push("conversational");
+          if (features.length > 0) {
+            stream2.write(`  ${import_chalk8.default.dim("Features:")}  ${import_chalk8.default.blue(features.join(", "))}
+`);
+          }
+        }
+      }
+    } else {
+      const idWidth = 32;
+      const nameWidth = 25;
+      const sizesWidth = 20;
+      const priceWidth = 15;
+      stream2.write(import_chalk8.default.dim("\u2500".repeat(idWidth + nameWidth + sizesWidth + priceWidth + 6)) + "\n");
+      stream2.write(
+        import_chalk8.default.bold(
+          "Model ID".padEnd(idWidth) + "  " + "Display Name".padEnd(nameWidth) + "  " + "Sizes".padEnd(sizesWidth) + "  " + "Price".padEnd(priceWidth)
+        ) + "\n"
+      );
+      stream2.write(import_chalk8.default.dim("\u2500".repeat(idWidth + nameWidth + sizesWidth + priceWidth + 6)) + "\n");
+      for (const model of providerModels) {
+        const sizes = model.supportedSizes.length > 2 ? model.supportedSizes.slice(0, 2).join(", ") + "..." : model.supportedSizes.join(", ");
+        stream2.write(
+          import_chalk8.default.green(model.modelId.padEnd(idWidth)) + "  " + import_chalk8.default.white(model.displayName.substring(0, nameWidth - 1).padEnd(nameWidth)) + "  " + import_chalk8.default.yellow(sizes.padEnd(sizesWidth)) + "  " + import_chalk8.default.cyan(formatImagePrice(model).padEnd(priceWidth)) + "\n"
+        );
+      }
+      stream2.write(import_chalk8.default.dim("\u2500".repeat(idWidth + nameWidth + sizesWidth + priceWidth + 6)) + "\n");
+    }
+    stream2.write("\n");
+  }
+}
+function renderSpeechTable(models, verbose, stream2) {
+  stream2.write(import_chalk8.default.bold.magenta("\u{1F3A4} Speech (TTS) Models\n"));
+  stream2.write(import_chalk8.default.dim("\u2500".repeat(80)) + "\n\n");
+  const grouped = /* @__PURE__ */ new Map();
+  for (const model of models) {
+    if (!grouped.has(model.provider)) {
+      grouped.set(model.provider, []);
+    }
+    grouped.get(model.provider).push(model);
+  }
+  for (const [provider, providerModels] of Array.from(grouped.entries()).sort()) {
+    const providerName = provider.charAt(0).toUpperCase() + provider.slice(1);
+    stream2.write(import_chalk8.default.bold.yellow(`${providerName}
+`));
+    if (verbose) {
+      for (const model of providerModels) {
+        stream2.write(import_chalk8.default.bold.green(`
+  ${model.modelId}
+`));
+        stream2.write(import_chalk8.default.dim("  " + "\u2500".repeat(60)) + "\n");
+        stream2.write(`  ${import_chalk8.default.dim("Name:")}    ${import_chalk8.default.white(model.displayName)}
+`);
+        stream2.write(`  ${import_chalk8.default.dim("Voices:")}  ${import_chalk8.default.yellow(model.voices.length.toString())} voices
+`);
+        if (model.voices.length <= 6) {
+          stream2.write(`            ${import_chalk8.default.dim(model.voices.join(", "))}
+`);
+        } else {
+          stream2.write(`            ${import_chalk8.default.dim(model.voices.slice(0, 6).join(", ") + "...")}
+`);
+        }
+        stream2.write(`  ${import_chalk8.default.dim("Formats:")} ${import_chalk8.default.yellow(model.formats.join(", "))}
+`);
+        stream2.write(`  ${import_chalk8.default.dim("Max Input:")} ${import_chalk8.default.yellow(model.maxInputLength.toString())} chars
+`);
+        stream2.write(`  ${import_chalk8.default.dim("Pricing:")} ${import_chalk8.default.cyan(formatSpeechPrice(model))}
+`);
+        if (model.features) {
+          const features = [];
+          if (model.features.multiSpeaker) features.push("multi-speaker");
+          if (model.features.voiceInstructions) features.push("voice-instructions");
+          if (model.features.languages) features.push(`${model.features.languages} languages`);
+          if (features.length > 0) {
+            stream2.write(`  ${import_chalk8.default.dim("Features:")} ${import_chalk8.default.blue(features.join(", "))}
+`);
+          }
+        }
+      }
+    } else {
+      const idWidth = 30;
+      const nameWidth = 28;
+      const voicesWidth = 12;
+      const priceWidth = 18;
+      stream2.write(import_chalk8.default.dim("\u2500".repeat(idWidth + nameWidth + voicesWidth + priceWidth + 6)) + "\n");
+      stream2.write(
+        import_chalk8.default.bold(
+          "Model ID".padEnd(idWidth) + "  " + "Display Name".padEnd(nameWidth) + "  " + "Voices".padEnd(voicesWidth) + "  " + "Price".padEnd(priceWidth)
+        ) + "\n"
+      );
+      stream2.write(import_chalk8.default.dim("\u2500".repeat(idWidth + nameWidth + voicesWidth + priceWidth + 6)) + "\n");
+      for (const model of providerModels) {
+        stream2.write(
+          import_chalk8.default.green(model.modelId.padEnd(idWidth)) + "  " + import_chalk8.default.white(model.displayName.substring(0, nameWidth - 1).padEnd(nameWidth)) + "  " + import_chalk8.default.yellow(`${model.voices.length} voices`.padEnd(voicesWidth)) + "  " + import_chalk8.default.cyan(formatSpeechPrice(model).padEnd(priceWidth)) + "\n"
+        );
+      }
+      stream2.write(import_chalk8.default.dim("\u2500".repeat(idWidth + nameWidth + voicesWidth + priceWidth + 6)) + "\n");
+    }
+    stream2.write("\n");
+  }
+}
+function formatImagePrice(model) {
+  if (model.pricing.perImage !== void 0) {
+    return `$${model.pricing.perImage.toFixed(2)}/img`;
+  }
+  if (model.pricing.bySize) {
+    const prices = Object.values(model.pricing.bySize);
+    const minPrice = Math.min(...prices.flatMap((p) => typeof p === "number" ? [p] : Object.values(p)));
+    const maxPrice = Math.max(...prices.flatMap((p) => typeof p === "number" ? [p] : Object.values(p)));
+    if (minPrice === maxPrice) {
+      return `$${minPrice.toFixed(2)}/img`;
+    }
+    return `$${minPrice.toFixed(2)}-${maxPrice.toFixed(2)}`;
+  }
+  return "varies";
+}
+function formatSpeechPrice(model) {
+  if (model.pricing.perCharacter !== void 0) {
+    const perMillion = model.pricing.perCharacter * 1e6;
+    return `$${perMillion.toFixed(0)}/1M chars`;
+  }
+  if (model.pricing.perMinute !== void 0) {
+    return `~$${model.pricing.perMinute.toFixed(2)}/min`;
+  }
+  return "varies";
+}
+function renderJSON(textModels, imageModels, speechModels, stream2) {
+  const output = {};
+  if (textModels.length > 0) {
+    output.textModels = textModels.map((model) => ({
       provider: model.provider,
       modelId: model.modelId,
       displayName: model.displayName,
@@ -11285,9 +12657,33 @@ function renderJSON(models, stream2) {
       knowledgeCutoff: model.knowledgeCutoff,
       features: model.features,
       metadata: model.metadata
-    })),
-    shortcuts: MODEL_ALIASES
-  };
+    }));
+    output.shortcuts = MODEL_ALIASES;
+  }
+  if (imageModels.length > 0) {
+    output.imageModels = imageModels.map((model) => ({
+      provider: model.provider,
+      modelId: model.modelId,
+      displayName: model.displayName,
+      supportedSizes: model.supportedSizes,
+      supportedQualities: model.supportedQualities,
+      maxImages: model.maxImages,
+      pricing: model.pricing,
+      features: model.features
+    }));
+  }
+  if (speechModels.length > 0) {
+    output.speechModels = speechModels.map((model) => ({
+      provider: model.provider,
+      modelId: model.modelId,
+      displayName: model.displayName,
+      voices: model.voices,
+      formats: model.formats,
+      maxInputLength: model.maxInputLength,
+      pricing: model.pricing,
+      features: model.features
+    }));
+  }
   stream2.write(JSON.stringify(output, null, 2) + "\n");
 }
 function formatTokens2(count) {
@@ -11300,7 +12696,7 @@ function formatTokens2(count) {
   }
 }
 function registerModelsCommand(program, env) {
-  program.command(COMMANDS.models).description("List all available LLM models with pricing and capabilities.").option("--provider <name>", "Filter by provider (openai, anthropic, gemini)").option("--format <format>", "Output format: table or json", "table").option("--verbose", "Show detailed model information", false).action(
+  program.command(COMMANDS.models).description("List available models with pricing and capabilities.").option("--provider <name>", "Filter by provider (openai, anthropic, gemini)").option("--format <format>", "Output format: table or json", "table").option("--verbose", "Show detailed model information", false).option("--text", "Show text/LLM models (default if no type specified)").option("--image", "Show image generation models").option("--speech", "Show speech/TTS models").option("--all", "Show all model types (text, image, speech)").action(
     (options) => executeAction(
       () => handleModelsCommand(options, env),
       env
@@ -11308,6 +12704,60 @@ function registerModelsCommand(program, env) {
   );
 }
 
+// src/cli/speech-command.ts
+var import_node_fs12 = require("fs");
+var DEFAULT_SPEECH_MODEL = "tts-1";
+var DEFAULT_VOICE = "nova";
+async function executeSpeech(textArg, options, env) {
+  const text = await resolvePrompt(textArg, env);
+  const client = env.createClient();
+  const model = options.model;
+  const voice = options.voice ?? DEFAULT_VOICE;
+  const speed = options.speed ? Number.parseFloat(options.speed) : void 0;
+  const stderrTTY = env.stderr.isTTY === true;
+  if (!options.quiet && stderrTTY) {
+    env.stderr.write(`${SUMMARY_PREFIX} Generating speech with ${model} (voice: ${voice})...
+`);
+  }
+  const result = await client.speech.generate({
+    model,
+    input: text,
+    voice,
+    responseFormat: options.format,
+    speed
+  });
+  const audioBuffer = Buffer.from(result.audio);
+  if (options.output) {
+    (0, import_node_fs12.writeFileSync)(options.output, audioBuffer);
+    if (!options.quiet) {
+      env.stderr.write(`${SUMMARY_PREFIX} Audio saved to ${options.output}
+`);
+    }
+  } else {
+    env.stdout.write(audioBuffer);
+  }
+  if (!options.quiet && stderrTTY) {
+    const parts = [
+      `${result.usage.characterCount} characters`,
+      `format: ${result.format}`
+    ];
+    if (result.cost !== void 0) {
+      parts.push(`cost: ${formatCost(result.cost)}`);
+    }
+    env.stderr.write(`${SUMMARY_PREFIX} ${parts.join(" | ")}
+`);
+  }
+}
+function registerSpeechCommand(program, env, config) {
+  program.command(COMMANDS.speech).description("Generate speech audio from text.").argument("[text]", "Text to convert to speech. If omitted, stdin is used when available.").option(
+    OPTION_FLAGS.model,
+    OPTION_DESCRIPTIONS.model,
+    config?.model ?? DEFAULT_SPEECH_MODEL
+  ).option(OPTION_FLAGS.voice, OPTION_DESCRIPTIONS.voice, config?.voice ?? DEFAULT_VOICE).option(OPTION_FLAGS.speechFormat, OPTION_DESCRIPTIONS.speechFormat, config?.format).option(OPTION_FLAGS.speechSpeed, OPTION_DESCRIPTIONS.speechSpeed, config?.speed?.toString()).option(OPTION_FLAGS.speechOutput, OPTION_DESCRIPTIONS.speechOutput, config?.output).option(OPTION_FLAGS.quiet, OPTION_DESCRIPTIONS.quiet, config?.quiet ?? false).action(
+    (text, options) => executeAction(() => executeSpeech(text, options, env), env)
+  );
+}
+
 // src/cli/environment.ts
 var import_node_readline = __toESM(require("readline"), 1);
 var import_chalk9 = __toESM(require("chalk"), 1);
@@ -11459,6 +12909,8 @@ function createProgram(env, config) {
   });
   registerCompleteCommand(program, env, config?.complete);
   registerAgentCommand(program, env, config?.agent);
+  registerImageCommand(program, env, config?.image);
+  registerSpeechCommand(program, env, config?.speech);
   registerModelsCommand(program, env);
   registerGadgetCommand(program, env);
   if (config) {