@huggingface/inference 3.0.0 → 3.0.1

package/README.md

@@ -42,15 +42,15 @@ const hf = new HfInference('your access token')
 
 Your access token should be kept private. If you need to protect it in front-end applications, we suggest setting up a proxy server that stores the access token.
 
-### Requesting third-party inference providers
+### Third-party inference providers
 
-You can request inference from third-party providers with the inference client.
+You can send inference requests to third-party providers with the inference client.
 
 Currently, we support the following providers: [Fal.ai](https://fal.ai), [Replicate](https://replicate.com), [Together](https://together.xyz) and [Sambanova](https://sambanova.ai).
 
-To make request to a third-party provider, you have to pass the `provider` parameter to the inference function. Make sure your request is authenticated with an access token.
+To send requests to a third-party provider, you have to pass the `provider` parameter to the inference function. Make sure your request is authenticated with an access token.
 ```ts
-const accessToken = "hf_..."; // Either a HF access token, or an API key from the 3rd party provider (Replicate in this example)
+const accessToken = "hf_..."; // Either a HF access token, or an API key from the third-party provider (Replicate in this example)
 
 const client = new HfInference(accessToken);
 await client.textToImage({
@@ -63,14 +63,19 @@ await client.textToImage({
 When authenticated with a Hugging Face access token, the request is routed through https://huggingface.co.
 When authenticated with a third-party provider key, the request is made directly against that provider's inference API.
 
-Only a subset of models are supported when requesting 3rd party providers. You can check the list of supported models per pipeline tasks here:
+Only a subset of models are supported when requesting third-party providers. You can check the list of supported models per pipeline tasks here:
 - [Fal.ai supported models](./src/providers/fal-ai.ts)
 - [Replicate supported models](./src/providers/replicate.ts)
 - [Sambanova supported models](./src/providers/sambanova.ts)
 - [Together supported models](./src/providers/together.ts)
 - [HF Inference API (serverless)](https://huggingface.co/models?inference=warm&sort=trending)
 
-#### Tree-shaking
+❗**Important note:** To be compatible, the third-party API must adhere to the "standard" shape API we expect on HF model pages for each pipeline task type. 
+This is not an issue for LLMs as everyone converged on the OpenAI API anyways, but can be more tricky for other tasks like "text-to-image" or "automatic-speech-recognition" where there exists no standard API. Let us know if any help is needed or if we can make things easier for you!
+
+👋**Want to add another provider?** Get in touch if you'd like to add support for another Inference provider, and/or request it on https://huggingface.co/spaces/huggingface/HuggingDiscussions/discussions/49
+
+### Tree-shaking
 
 You can import the functions you need directly from the module instead of using the `HfInference` class.
 

package/dist/index.cjs

@@ -107,7 +107,15 @@ var FAL_AI_API_BASE_URL = "https://fal.run";
 var FAL_AI_SUPPORTED_MODEL_IDS = {
   "text-to-image": {
     "black-forest-labs/FLUX.1-schnell": "fal-ai/flux/schnell",
-    "black-forest-labs/FLUX.1-dev": "fal-ai/flux/dev"
+    "black-forest-labs/FLUX.1-dev": "fal-ai/flux/dev",
+    "playgroundai/playground-v2.5-1024px-aesthetic": "fal-ai/playground-v25",
+    "ByteDance/SDXL-Lightning": "fal-ai/lightning-models",
+    "PixArt-alpha/PixArt-Sigma-XL-2-1024-MS": "fal-ai/pixart-sigma",
+    "stabilityai/stable-diffusion-3-medium": "fal-ai/stable-diffusion-v3-medium",
+    "Warlord-K/Sana-1024": "fal-ai/sana",
+    "fal/AuraFlow-v0.2": "fal-ai/aura-flow",
+    "stabilityai/stable-diffusion-3.5-large": "fal-ai/stable-diffusion-v35-large",
+    "Kwai-Kolors/Kolors": "fal-ai/kolors"
   },
   "automatic-speech-recognition": {
     "openai/whisper-large-v3": "fal-ai/whisper"
@@ -120,10 +128,10 @@ var REPLICATE_SUPPORTED_MODEL_IDS = {
   "text-to-image": {
     "black-forest-labs/FLUX.1-schnell": "black-forest-labs/flux-schnell",
     "ByteDance/SDXL-Lightning": "bytedance/sdxl-lightning-4step:5599ed30703defd1d160a25a63321b4dec97101d98b4674bcc56e41f62f35637"
+  },
+  "text-to-speech": {
+    "OuteAI/OuteTTS-0.3-500M": "jbilcke/oute-tts:39a59319327b27327fa3095149c5a746e7f2aee18c75055c3368237a6503cd26"
   }
-  // "text-to-speech": {
-  // 	"SWivid/F5-TTS": "x-lance/f5-tts:87faf6dd7a692dd82043f662e76369cab126a2cf1937e25a9d41e0b834fd230e"
-  // },
 };
 
 // src/providers/sambanova.ts
@@ -159,6 +167,8 @@ var TOGETHER_SUPPORTED_MODEL_IDS = {
   },
   conversational: {
     "databricks/dbrx-instruct": "databricks/dbrx-instruct",
+    "deepseek-ai/DeepSeek-R1": "deepseek-ai/DeepSeek-R1",
+    "deepseek-ai/DeepSeek-V3": "deepseek-ai/DeepSeek-V3",
     "deepseek-ai/deepseek-llm-67b-chat": "deepseek-ai/deepseek-llm-67b-chat",
     "google/gemma-2-9b-it": "google/gemma-2-9b-it",
     "google/gemma-2b-it": "google/gemma-2-27b-it",
@@ -204,7 +214,8 @@ function isUrl(modelOrUrl) {
 var HF_HUB_INFERENCE_PROXY_TEMPLATE = `${HF_HUB_URL}/api/inference-proxy/{{PROVIDER}}`;
 var tasks = null;
 async function makeRequestOptions(args, options) {
-  const { accessToken, endpointUrl, provider: maybeProvider, model: maybeModel, ...otherArgs } = args;
+  const { accessToken, endpointUrl, provider: maybeProvider, model: maybeModel, ...remainingArgs } = args;
+  let otherArgs = remainingArgs;
   const provider = maybeProvider ?? "hf-inference";
   const { forceTask, includeCredentials, taskHint, wait_for_model, use_cache, dont_load_model, chatCompletion: chatCompletion2 } = options ?? {};
   if (endpointUrl && provider !== "hf-inference") {
@@ -263,9 +274,9 @@ async function makeRequestOptions(args, options) {
   } else if (includeCredentials === true) {
     credentials = "include";
   }
-  if (provider === "replicate" && model.includes(":")) {
-    const version = model.split(":")[1];
-    otherArgs.version = version;
+  if (provider === "replicate") {
+    const version = model.includes(":") ? model.split(":")[1] : void 0;
+    otherArgs = { input: otherArgs, version };
   }
   const info = {
     headers,
@@ -638,6 +649,19 @@ async function textToSpeech(args, options) {
     ...options,
     taskHint: "text-to-speech"
   });
+  if (res && typeof res === "object") {
+    if ("output" in res) {
+      if (typeof res.output === "string") {
+        const urlResponse = await fetch(res.output);
+        const blob = await urlResponse.blob();
+        return blob;
+      } else if (Array.isArray(res.output)) {
+        const urlResponse = await fetch(res.output[0]);
+        const blob = await urlResponse.blob();
+        return blob;
+      }
+    }
+  }
   const isValidOutput = res && res instanceof Blob;
   if (!isValidOutput) {
     throw new InferenceOutputError("Expected Blob");
@@ -719,10 +743,10 @@ async function objectDetection(args, options) {
 async function textToImage(args, options) {
   if (args.provider === "together" || args.provider === "fal-ai") {
     args.prompt = args.inputs;
-    args.inputs = "";
+    delete args.inputs;
     args.response_format = "base64";
   } else if (args.provider === "replicate") {
-    args.input = { prompt: args.inputs };
+    args.prompt = args.inputs;
     delete args.inputs;
   }
   const res = await request(args, {

package/dist/index.js

@@ -49,7 +49,15 @@ var FAL_AI_API_BASE_URL = "https://fal.run";
 var FAL_AI_SUPPORTED_MODEL_IDS = {
   "text-to-image": {
     "black-forest-labs/FLUX.1-schnell": "fal-ai/flux/schnell",
-    "black-forest-labs/FLUX.1-dev": "fal-ai/flux/dev"
+    "black-forest-labs/FLUX.1-dev": "fal-ai/flux/dev",
+    "playgroundai/playground-v2.5-1024px-aesthetic": "fal-ai/playground-v25",
+    "ByteDance/SDXL-Lightning": "fal-ai/lightning-models",
+    "PixArt-alpha/PixArt-Sigma-XL-2-1024-MS": "fal-ai/pixart-sigma",
+    "stabilityai/stable-diffusion-3-medium": "fal-ai/stable-diffusion-v3-medium",
+    "Warlord-K/Sana-1024": "fal-ai/sana",
+    "fal/AuraFlow-v0.2": "fal-ai/aura-flow",
+    "stabilityai/stable-diffusion-3.5-large": "fal-ai/stable-diffusion-v35-large",
+    "Kwai-Kolors/Kolors": "fal-ai/kolors"
   },
   "automatic-speech-recognition": {
     "openai/whisper-large-v3": "fal-ai/whisper"
@@ -62,10 +70,10 @@ var REPLICATE_SUPPORTED_MODEL_IDS = {
   "text-to-image": {
     "black-forest-labs/FLUX.1-schnell": "black-forest-labs/flux-schnell",
     "ByteDance/SDXL-Lightning": "bytedance/sdxl-lightning-4step:5599ed30703defd1d160a25a63321b4dec97101d98b4674bcc56e41f62f35637"
+  },
+  "text-to-speech": {
+    "OuteAI/OuteTTS-0.3-500M": "jbilcke/oute-tts:39a59319327b27327fa3095149c5a746e7f2aee18c75055c3368237a6503cd26"
   }
-  // "text-to-speech": {
-  // 	"SWivid/F5-TTS": "x-lance/f5-tts:87faf6dd7a692dd82043f662e76369cab126a2cf1937e25a9d41e0b834fd230e"
-  // },
 };
 
 // src/providers/sambanova.ts
@@ -101,6 +109,8 @@ var TOGETHER_SUPPORTED_MODEL_IDS = {
   },
   conversational: {
     "databricks/dbrx-instruct": "databricks/dbrx-instruct",
+    "deepseek-ai/DeepSeek-R1": "deepseek-ai/DeepSeek-R1",
+    "deepseek-ai/DeepSeek-V3": "deepseek-ai/DeepSeek-V3",
     "deepseek-ai/deepseek-llm-67b-chat": "deepseek-ai/deepseek-llm-67b-chat",
     "google/gemma-2-9b-it": "google/gemma-2-9b-it",
     "google/gemma-2b-it": "google/gemma-2-27b-it",
@@ -146,7 +156,8 @@ function isUrl(modelOrUrl) {
 var HF_HUB_INFERENCE_PROXY_TEMPLATE = `${HF_HUB_URL}/api/inference-proxy/{{PROVIDER}}`;
 var tasks = null;
 async function makeRequestOptions(args, options) {
-  const { accessToken, endpointUrl, provider: maybeProvider, model: maybeModel, ...otherArgs } = args;
+  const { accessToken, endpointUrl, provider: maybeProvider, model: maybeModel, ...remainingArgs } = args;
+  let otherArgs = remainingArgs;
   const provider = maybeProvider ?? "hf-inference";
   const { forceTask, includeCredentials, taskHint, wait_for_model, use_cache, dont_load_model, chatCompletion: chatCompletion2 } = options ?? {};
   if (endpointUrl && provider !== "hf-inference") {
@@ -205,9 +216,9 @@ async function makeRequestOptions(args, options) {
   } else if (includeCredentials === true) {
     credentials = "include";
   }
-  if (provider === "replicate" && model.includes(":")) {
-    const version = model.split(":")[1];
-    otherArgs.version = version;
+  if (provider === "replicate") {
+    const version = model.includes(":") ? model.split(":")[1] : void 0;
+    otherArgs = { input: otherArgs, version };
   }
   const info = {
     headers,
@@ -580,6 +591,19 @@ async function textToSpeech(args, options) {
     ...options,
     taskHint: "text-to-speech"
   });
+  if (res && typeof res === "object") {
+    if ("output" in res) {
+      if (typeof res.output === "string") {
+        const urlResponse = await fetch(res.output);
+        const blob = await urlResponse.blob();
+        return blob;
+      } else if (Array.isArray(res.output)) {
+        const urlResponse = await fetch(res.output[0]);
+        const blob = await urlResponse.blob();
+        return blob;
+      }
+    }
+  }
   const isValidOutput = res && res instanceof Blob;
   if (!isValidOutput) {
     throw new InferenceOutputError("Expected Blob");
@@ -661,10 +685,10 @@ async function objectDetection(args, options) {
 async function textToImage(args, options) {
   if (args.provider === "together" || args.provider === "fal-ai") {
     args.prompt = args.inputs;
-    args.inputs = "";
+    delete args.inputs;
     args.response_format = "base64";
   } else if (args.provider === "replicate") {
-    args.input = { prompt: args.inputs };
+    args.prompt = args.inputs;
     delete args.inputs;
   }
   const res = await request(args, {

package/dist/src/lib/makeRequestOptions.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"makeRequestOptions.d.ts","sourceRoot":"","sources":["../../../src/lib/makeRequestOptions.ts"],"names":[],"mappings":"AAOA,OAAO,KAAK,EAAE,aAAa,EAAE,OAAO,EAAE,WAAW,EAAE,MAAM,UAAU,CAAC;AAWpE;;GAEG;AACH,wBAAsB,kBAAkB,CACvC,IAAI,EAAE,WAAW,GAAG;IACnB,IAAI,CAAC,EAAE,IAAI,GAAG,WAAW,CAAC;IAC1B,MAAM,CAAC,EAAE,OAAO,CAAC;CACjB,EACD,OAAO,CAAC,EAAE,OAAO,GAAG;IACnB,yFAAyF;IACzF,SAAS,CAAC,EAAE,MAAM,GAAG,aAAa,CAAC;IACnC,sCAAsC;IACtC,QAAQ,CAAC,EAAE,aAAa,CAAC;IACzB,cAAc,CAAC,EAAE,OAAO,CAAC;CACzB,GACC,OAAO,CAAC;IAAE,GAAG,EAAE,MAAM,CAAC;IAAC,IAAI,EAAE,WAAW,CAAA;CAAE,CAAC,CA+G7C"}
+{"version":3,"file":"makeRequestOptions.d.ts","sourceRoot":"","sources":["../../../src/lib/makeRequestOptions.ts"],"names":[],"mappings":"AAOA,OAAO,KAAK,EAAE,aAAa,EAAE,OAAO,EAAE,WAAW,EAAE,MAAM,UAAU,CAAC;AAWpE;;GAEG;AACH,wBAAsB,kBAAkB,CACvC,IAAI,EAAE,WAAW,GAAG;IACnB,IAAI,CAAC,EAAE,IAAI,GAAG,WAAW,CAAC;IAC1B,MAAM,CAAC,EAAE,OAAO,CAAC;CACjB,EACD,OAAO,CAAC,EAAE,OAAO,GAAG;IACnB,yFAAyF;IACzF,SAAS,CAAC,EAAE,MAAM,GAAG,aAAa,CAAC;IACnC,sCAAsC;IACtC,QAAQ,CAAC,EAAE,aAAa,CAAC;IACzB,cAAc,CAAC,EAAE,OAAO,CAAC;CACzB,GACC,OAAO,CAAC;IAAE,GAAG,EAAE,MAAM,CAAC;IAAC,IAAI,EAAE,WAAW,CAAA;CAAE,CAAC,CAiH7C"}

package/dist/src/providers/fal-ai.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"fal-ai.d.ts","sourceRoot":"","sources":["../../../src/providers/fal-ai.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,SAAS,CAAC;AAE/C,eAAO,MAAM,mBAAmB,oBAAoB,CAAC;AAErD,KAAK,OAAO,GAAG,MAAM,CAAC;AAEtB,eAAO,MAAM,0BAA0B,EAAE,eAAe,CAAC,OAAO,CAQ/D,CAAC"}
+{"version":3,"file":"fal-ai.d.ts","sourceRoot":"","sources":["../../../src/providers/fal-ai.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,SAAS,CAAC;AAE/C,eAAO,MAAM,mBAAmB,oBAAoB,CAAC;AAErD,KAAK,OAAO,GAAG,MAAM,CAAC;AAEtB,eAAO,MAAM,0BAA0B,EAAE,eAAe,CAAC,OAAO,CAgB/D,CAAC"}

package/dist/src/providers/together.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"together.d.ts","sourceRoot":"","sources":["../../../src/providers/together.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,SAAS,CAAC;AAE/C,eAAO,MAAM,qBAAqB,6BAA6B,CAAC;AAEhE;;GAEG;AACH,KAAK,UAAU,GAAG,MAAM,CAAC;AAEzB;;GAEG;AACH,eAAO,MAAM,4BAA4B,EAAE,eAAe,CAAC,UAAU,CA6CpE,CAAC"}
+{"version":3,"file":"together.d.ts","sourceRoot":"","sources":["../../../src/providers/together.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,SAAS,CAAC;AAE/C,eAAO,MAAM,qBAAqB,6BAA6B,CAAC;AAEhE;;GAEG;AACH,KAAK,UAAU,GAAG,MAAM,CAAC;AAEzB;;GAEG;AACH,eAAO,MAAM,4BAA4B,EAAE,eAAe,CAAC,UAAU,CA+CpE,CAAC"}

package/dist/src/tasks/audio/textToSpeech.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"textToSpeech.d.ts","sourceRoot":"","sources":["../../../../src/tasks/audio/textToSpeech.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,aAAa,CAAC;AAGrD,MAAM,MAAM,gBAAgB,GAAG,QAAQ,GAAG;IACzC;;OAEG;IACH,MAAM,EAAE,MAAM,CAAC;CACf,CAAC;AAEF,MAAM,MAAM,kBAAkB,GAAG,IAAI,CAAC;AAEtC;;;GAGG;AACH,wBAAsB,YAAY,CAAC,IAAI,EAAE,gBAAgB,EAAE,OAAO,CAAC,EAAE,OAAO,GAAG,OAAO,CAAC,kBAAkB,CAAC,CAUzG"}
+{"version":3,"file":"textToSpeech.d.ts","sourceRoot":"","sources":["../../../../src/tasks/audio/textToSpeech.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,aAAa,CAAC;AAGrD,MAAM,MAAM,gBAAgB,GAAG,QAAQ,GAAG;IACzC;;OAEG;IACH,MAAM,EAAE,MAAM,CAAC;CACf,CAAC;AAEF,MAAM,MAAM,kBAAkB,GAAG,IAAI,CAAC;AAItC;;;GAGG;AACH,wBAAsB,YAAY,CAAC,IAAI,EAAE,gBAAgB,EAAE,OAAO,CAAC,EAAE,OAAO,GAAG,OAAO,CAAC,kBAAkB,CAAC,CAuBzG"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@huggingface/inference",
-  "version": "3.0.0",
+  "version": "3.0.1",
   "packageManager": "pnpm@8.10.5",
   "license": "MIT",
   "author": "Tim Mikeladze <tim.mikeladze@gmail.com>",
@@ -39,7 +39,7 @@
   },
   "type": "module",
   "dependencies": {
-    "@huggingface/tasks": "^0.13.16"
+    "@huggingface/tasks": "^0.13.17"
   },
   "devDependencies": {
     "@types/node": "18.13.0"

package/src/lib/makeRequestOptions.ts

@@ -32,7 +32,8 @@ export async function makeRequestOptions(
 		chatCompletion?: boolean;
 	}
 ): Promise<{ url: string; info: RequestInit }> {
-	const { accessToken, endpointUrl, provider: maybeProvider, model: maybeModel, ...otherArgs } = args;
+	const { accessToken, endpointUrl, provider: maybeProvider, model: maybeModel, ...remainingArgs } = args;
+	let otherArgs = remainingArgs;
 	const provider = maybeProvider ?? "hf-inference";
 
 	const { forceTask, includeCredentials, taskHint, wait_for_model, use_cache, dont_load_model, chatCompletion } =
@@ -120,12 +121,13 @@ export async function makeRequestOptions(
 		credentials = "include";
 	}
 
-	/*
+	/**
+	 * Replicate models wrap all inputs inside { input: ... }
 	 * Versioned Replicate models in the format `owner/model:version` expect the version in the body
 	 */
-	if (provider === "replicate" && model.includes(":")) {
-		const version = model.split(":")[1];
-		(otherArgs as typeof otherArgs & { version: string }).version = version;
+	if (provider === "replicate") {
+		const version = model.includes(":") ? model.split(":")[1] : undefined;
+		(otherArgs as unknown) = { input: otherArgs, version };
 	}
 
 	const info: RequestInit = {

package/src/providers/fal-ai.ts

@@ -8,6 +8,14 @@ export const FAL_AI_SUPPORTED_MODEL_IDS: ProviderMapping<FalAiId> = {
 	"text-to-image": {
 		"black-forest-labs/FLUX.1-schnell": "fal-ai/flux/schnell",
 		"black-forest-labs/FLUX.1-dev": "fal-ai/flux/dev",
+		"playgroundai/playground-v2.5-1024px-aesthetic": "fal-ai/playground-v25",
+		"ByteDance/SDXL-Lightning": "fal-ai/lightning-models",
+		"PixArt-alpha/PixArt-Sigma-XL-2-1024-MS": "fal-ai/pixart-sigma",
+		"stabilityai/stable-diffusion-3-medium": "fal-ai/stable-diffusion-v3-medium",
+		"Warlord-K/Sana-1024": "fal-ai/sana",
+		"fal/AuraFlow-v0.2": "fal-ai/aura-flow",
+		"stabilityai/stable-diffusion-3.5-large": "fal-ai/stable-diffusion-v35-large",
+		"Kwai-Kolors/Kolors": "fal-ai/kolors",
 	},
 	"automatic-speech-recognition": {
 		"openai/whisper-large-v3": "fal-ai/whisper",

package/src/providers/replicate.ts

@@ -10,7 +10,7 @@ export const REPLICATE_SUPPORTED_MODEL_IDS: ProviderMapping<ReplicateId> = {
 		"ByteDance/SDXL-Lightning":
 			"bytedance/sdxl-lightning-4step:5599ed30703defd1d160a25a63321b4dec97101d98b4674bcc56e41f62f35637",
 	},
-	// "text-to-speech": {
-	// 	"SWivid/F5-TTS": "x-lance/f5-tts:87faf6dd7a692dd82043f662e76369cab126a2cf1937e25a9d41e0b834fd230e"
-	// },
+	"text-to-speech": {
+		"OuteAI/OuteTTS-0.3-500M": "jbilcke/oute-tts:39a59319327b27327fa3095149c5a746e7f2aee18c75055c3368237a6503cd26",
+	},
 };

package/src/providers/together.ts

@@ -21,6 +21,8 @@ export const TOGETHER_SUPPORTED_MODEL_IDS: ProviderMapping<TogetherId> = {
 	},
 	conversational: {
 		"databricks/dbrx-instruct": "databricks/dbrx-instruct",
+		"deepseek-ai/DeepSeek-R1": "deepseek-ai/DeepSeek-R1",
+		"deepseek-ai/DeepSeek-V3": "deepseek-ai/DeepSeek-V3",
 		"deepseek-ai/deepseek-llm-67b-chat": "deepseek-ai/deepseek-llm-67b-chat",
 		"google/gemma-2-9b-it": "google/gemma-2-9b-it",
 		"google/gemma-2b-it": "google/gemma-2-27b-it",

package/src/tasks/audio/textToSpeech.ts

@@ -10,16 +10,31 @@ export type TextToSpeechArgs = BaseArgs & {
 };
 
 export type TextToSpeechOutput = Blob;
-
+interface OutputUrlTextToSpeechGeneration {
+	output: string | string[];
+}
 /**
  * This task synthesize an audio of a voice pronouncing a given text.
  * Recommended model: espnet/kan-bayashi_ljspeech_vits
  */
 export async function textToSpeech(args: TextToSpeechArgs, options?: Options): Promise<TextToSpeechOutput> {
-	const res = await request<TextToSpeechOutput>(args, {
+	const res = await request<TextToSpeechOutput | OutputUrlTextToSpeechGeneration>(args, {
 		...options,
 		taskHint: "text-to-speech",
 	});
+	if (res && typeof res === "object") {
+		if ("output" in res) {
+			if (typeof res.output === "string") {
+				const urlResponse = await fetch(res.output);
+				const blob = await urlResponse.blob();
+				return blob;
+			} else if (Array.isArray(res.output)) {
+				const urlResponse = await fetch(res.output[0]);
+				const blob = await urlResponse.blob();
+				return blob;
+			}
+		}
+	}
 	const isValidOutput = res && res instanceof Blob;
 	if (!isValidOutput) {
 		throw new InferenceOutputError("Expected Blob");

package/src/tasks/cv/textToImage.ts

@@ -59,10 +59,10 @@ interface OutputUrlImageGeneration {
 export async function textToImage(args: TextToImageArgs, options?: Options): Promise<TextToImageOutput> {
 	if (args.provider === "together" || args.provider === "fal-ai") {
 		args.prompt = args.inputs;
-		args.inputs = "";
+		delete (args as unknown as { inputs: unknown }).inputs;
 		args.response_format = "base64";
 	} else if (args.provider === "replicate") {
-		args.input = { prompt: args.inputs };
+		args.prompt = args.inputs;
 		delete (args as unknown as { inputs: unknown }).inputs;
 	}
 	const res = await request<TextToImageOutput | Base64ImageGeneration | OutputUrlImageGeneration>(args, {