# Cartesia MCP for AI Agents AI Agent Connect > Cartesia (Voice AI) lets your AI agent speak with human-like voices, clone specific voices from short clips, and transcribe audio in real time. It handles the heavy lifting of high-quality speech synthesis and recognition so you don't have to build the audio infrastructure yourself. ## Overview - **Category:** ai-frontier - **Price:** Free - **Endpoint:** https://edge.vinkius.com/vk_preview_fqUkC4o6TEWq9kAU8DlY8MAb7pfJxo3R952S85p7/ai-agent-connect - **Tags:** text-to-speech, speech-to-text, voice-synthesis, low-latency, ai-voice, audio-streaming ## Description Cartesia connects your AI agent to high-performance voice synthesis and speech recognition. Instead of dealing with robotic tones or long delays, you get audio that sounds like a real person. This makes conversations feel fluid and natural, which is a huge deal for customer-facing bots or content creation tools. You don't have to worry about the backend work of streaming audio or managing complex infrastructure. You just tell your agent to speak, and it happens. If you're looking for a way to put this into production, the Vinkius catalog makes it easy to plug this into your existing setup. You get to focus on the logic of the conversation while the Connector handles the actual sound. Whether you're building a customer service bot that needs to sound empathetic or a content tool that needs to turn text into high-quality voiceovers, this is the bridge between text and sound. It handles everything from the initial transcription of a voice note to the final synthesis of a multi-language response. You can also create custom voices from short clips or manage how specific terms are spoken. This means your agent can handle technical jargon or brand names without stumbling. It's built for speed, so you get responses that keep up with the pace of a real human conversation. ## Tools ### get_usage_credits Check how many credits you have left. Use this to monitor your costs during heavy testing. ### localize_voice Change a voice's language or dialect. This is how you make a voice sound local to a specific region. ### clone_voice Create a new voice from a 5-second audio clip. This is the fastest way to get a custom voice model. ### create_pronunciation_dict Make a list of how specific words should sound. Use this for technical jargon or brand names. ### delete_pronunciation_dict Remove a pronunciation dictionary you no longer need. Keep your library clean. ### delete_voice Permanently remove a voice from your account. Use this to manage your storage and limits. ### generate_access_token Get a short-lived token for client-side requests. This helps keep your main API key secure. ### get_agent Pull the details of a specific voice agent. Useful for checking agent configurations on the fly. ### infill_bytes Generate audio to bridge two existing audio segments. It makes your voice clips sound continuous. ### list_agent_calls See the history of calls and transcripts for a specific agent. It's great for debugging conversation flows. ### list_agents See all the voice agents you've created. Use this to pick the right one for a task. ### list_pronunciation_dicts View all your custom pronunciation rules. This helps you manage how your AI handles specific words. ### list_voices See every voice available in your account. Use this to browse your library of voices. ### get_voice Get the specific details of a voice model. Use this to check IDs before you start a project. ### update_voice Change the metadata for an existing voice. Use this to keep your voice library organized. ### stt_batch Transcribe a whole folder of audio files into text at once. It's the fastest way to handle bulk transcription. ### tts_bytes Turn text into a downloadable audio file. Use this when you need a static file instead of a stream. ### tts_sse Stream text-to-speech audio as it's being generated. This is the best way to achieve low-latency conversation. ### update_pronunciation_dict Edit your existing pronunciation rules. Use this to fix how the AI says a specific term. ### voice_changer_bytes Swap the voice of an existing audio clip. It keeps the original tone but changes the speaker. ## Prompt Examples **Prompt:** ``` Can you tell me what voices are available in my account? ``` **Response:** ``` I've checked your Cartesia account. You have access to: * **Baritone** (ID: 79a045e3...) * **British Lady** (ID: 123x-456y...) * **3 custom cloned voices** Which one would you like to use for this project? ``` **Prompt:** ``` Generate a WAV file saying 'Hello, how can I help you today?' using my first custom voice. ``` **Response:** ``` Generating audio... I've synthesized the text using your custom voice model. Your WAV file is ready for download below: [Download Link: voice_output_01.wav] ``` **Prompt:** ``` How many credits do I have left? ``` **Response:** ``` You currently have **45,200 credits** remaining in your Cartesia account. Your last billing cycle refreshed on the 1st of the month. ``` ## Capabilities ### Clone a specific human voice Create a custom voice model from a 5-second audio clip. ### Turn text into high-fidelity audio Generate audio bytes for immediate playback or streaming. ### Stream audio for real-time apps Use Server-Sent Events to deliver audio with minimal latency. ### Transcribe audio into text Convert audio files into text across multiple languages. ### Create custom pronunciation rules Manage how the AI handles technical terms or brand names. ### Monitor agent call logs Keep track of call history and credit usage statistics. ### Adapt voices to different dialects Adjust a voice to sound local to a specific language or region. ## Use Cases ### Low-latency phone bots A developer wants a phone bot that doesn't lag. They use tts_sse to stream responses instantly for a natural flow. ### Custom voice narrations A YouTuber wants to clone their own voice. They use clone_voice and tts_bytes to generate narrations for their videos. ### Bulk meeting transcription A company needs to transcribe 100 hours of meetings. They use stt_batch to get text files quickly for their records. ### Global content localization A brand wants a specific accent for a global ad. They use localize_voice to adapt a voice to a new region. ## Benefits - Get human-like speech with tts_sse that feels like a real conversation. - Build custom brand voices quickly using clone_voice with just 5 seconds of audio. - Handle complex technical terms accurately with create_pronunciation_dict for better clarity. - Manage your costs and limits easily by checking get_usage_credits regularly. - Automate high-volume transcription tasks using stt_batch for faster data processing. - Create smooth transitions between audio clips using infill_bytes for professional-sounding outputs. ## How It Works The bottom line is you get human-sounding speech and transcription without building your own audio engine. 1. Connect your Cartesia API key to the Connector in your AI client. 2. Send a text string or audio file to your agent. 3. Receive back high-quality audio bytes or a text transcription. ## Frequently Asked Questions **Does Cartesia Voice AI support real-time conversation?** Yes, it supports real-time speech. By using the streaming tools, your AI agent can respond with almost zero delay, making it perfect for voice bots. **How do I make the AI say technical words correctly?** You can create a custom dictionary of words. This tells the AI exactly how to pronounce specific terms, brand names, or industry jargon. **Can I use my own voice for an AI agent?** Yes, you can clone a voice from a short audio clip. This allows you to create a custom voice model that sounds just like you or anyone else. **Is there a way to transcribe multiple files at once?** Yes, the Connector includes a batch transcription tool. You can send a whole list of audio files and get the text back in one go. **How do I see if I'm running out of credits?** You can ask your agent to check your usage stats. It will pull your current credit balance and tell you how much you have left for the month. **Can the AI speak in different accents?** Yes, the Connector allows you to adapt a voice to a new language or dialect. This is great for making your content feel local to different countries. **Can I generate audio in different formats like MP3 or WAV?** Yes. Using the `tts_bytes` tool, you can specify the `output_format_container` as 'mp3', 'wav', or 'raw', and configure the sample rate and encoding to match your needs. **How do I transcribe an existing audio file to text?** Use the `stt_batch` tool. Provide the base64 encoded audio file, specify the model (e.g., 'ink-whisper'), and the language code to receive a full transcription. **Is it possible to clone a voice using this integration?** Absolutely. The `clone_voice` tool allows you to create a new voice model by uploading a short (approx. 5s) base64 encoded audio clip.