Input

Video Player is loading.

Current Time 00:00:000

Duration 00:00:000

Loaded: 0%

Stream Type LIVE

Remaining Time 00:00:000

text

string

Shift + Return to add a new line

Then I would never talk to that person about boa constrictors, or primeval forests, or stars. I would bring myself down to his level.Then I would never talk to that person about boa constrictors, or primeval forests, or stars. I would bring myself down to his level.

Text to synthesize

Default: "Wow! That was an incredible firework display."

audio_prompt_path

file

Reference audio file to clone

exaggeration

number

Controls how expressive or exaggerated the speech sounds; higher values increase emotional intensity.

Default: 0.5

cfg_weight

number

Balances text fidelity and creativity; higher values make speech closer to the input text.

Default: 0.5

temperature

number

Adjusts randomness in speech generation; higher values produce more varied and natural output.

Default: 0.8

Run this model in Node.js with one line of code:

npx create-replicate --model=thomcle/chatterbox-tts

or set up a project from scratch

Install Replicate’s Node.js client library:

npm install replicate

Set the REPLICATE_API_TOKEN environment variable:

export REPLICATE_API_TOKEN=<paste-your-token-here>

Find your API token in your account settings.

Import and set up the client:

import Replicate from "replicate";
import fs from "node:fs";

const replicate = new Replicate({
  auth: process.env.REPLICATE_API_TOKEN,
});

Run thomcle/chatterbox-tts using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

const output = await replicate.run(
  "thomcle/chatterbox-tts:3f5f9c195086737dda710bf504330f71e786d0a361b505e377c8b10122af9d32",
  {
    input: {
      text: "Then I would never talk to that person about boa constrictors, or primeval forests, or stars. I would bring myself down to his level.",
      cfg_weight: 0.5,
      temperature: 0.8,
      exaggeration: 0.5,
      audio_prompt_path: "https://maskgct.github.io/audios/celeb_samples/rick_0.wav"
    }
  }
);

// To access the file URL:
console.log(output.url()); //=> "http://example.com"

// To write the file to disk:
fs.writeFile("my-image.png", output);

To learn more, take a look at the guide on getting started with Node.js.

Install Replicate’s Python client library:

pip install replicate

Set the REPLICATE_API_TOKEN environment variable:

export REPLICATE_API_TOKEN=<paste-your-token-here>

Find your API token in your account settings.

Import the client:

import replicate

Run thomcle/chatterbox-tts using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

output = replicate.run(
    "thomcle/chatterbox-tts:3f5f9c195086737dda710bf504330f71e786d0a361b505e377c8b10122af9d32",
    input={
        "text": "Then I would never talk to that person about boa constrictors, or primeval forests, or stars. I would bring myself down to his level.",
        "cfg_weight": 0.5,
        "temperature": 0.8,
        "exaggeration": 0.5,
        "audio_prompt_path": "https://maskgct.github.io/audios/celeb_samples/rick_0.wav"
    }
)

# To access the file URL:
print(output.url())
#=> "http://example.com"

# To write the file to disk:
with open("my-image.png", "wb") as file:
    file.write(output.read())

To learn more, take a look at the guide on getting started with Python.

Set the REPLICATE_API_TOKEN environment variable:

export REPLICATE_API_TOKEN=<paste-your-token-here>

Find your API token in your account settings.

Run thomcle/chatterbox-tts using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

curl -s -X POST \
  -H "Authorization: Bearer $REPLICATE_API_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Prefer: wait" \
  -d $'{
    "version": "thomcle/chatterbox-tts:3f5f9c195086737dda710bf504330f71e786d0a361b505e377c8b10122af9d32",
    "input": {
      "text": "Then I would never talk to that person about boa constrictors, or primeval forests, or stars. I would bring myself down to his level.",
      "cfg_weight": 0.5,
      "temperature": 0.8,
      "exaggeration": 0.5,
      "audio_prompt_path": "https://maskgct.github.io/audios/celeb_samples/rick_0.wav"
    }
  }' \
  https://api.replicate.com/v1/predictions

To learn more, take a look at Replicate’s HTTP API reference docs.

Output

Video Player is loading.

Current Time 00:00:000

Duration 00:00:000

Loaded: 0%

Stream Type LIVE

Remaining Time 00:00:000

{
  "completed_at": "2025-05-30T13:56:20.810154Z",
  "created_at": "2025-05-30T13:56:13.562000Z",
  "data_removed": false,
  "error": null,
  "id": "k0jzgr4yf9rga0cq45x9w5zfn8",
  "input": {
    "text": "Then I would never talk to that person about boa constrictors, or primeval forests, or stars. I would bring myself down to his level.",
    "cfg_weight": 0.5,
    "temperature": 0.8,
    "exaggeration": 0.5,
    "audio_prompt_path": "https://maskgct.github.io/audios/celeb_samples/rick_0.wav"
  },
  "logs": "WARNING:root:Reference mel length is not equal to 2 * reference token length.\nSampling:   0%|          | 0/1000 [00:00<?, ?it/s]\nSampling:   0%|          | 2/1000 [00:00<00:58, 17.12it/s]\nSampling:   1%|          | 6/1000 [00:00<00:39, 25.40it/s]\nSampling:   1%|          | 10/1000 [00:00<00:34, 28.29it/s]\nSampling:   1%|▏         | 14/1000 [00:00<00:33, 29.14it/s]\nSampling:   2%|▏         | 18/1000 [00:00<00:33, 29.71it/s]\nSampling:   2%|▏         | 22/1000 [00:00<00:31, 30.60it/s]\nSampling:   3%|▎         | 26/1000 [00:00<00:31, 31.15it/s]\nSampling:   3%|▎         | 30/1000 [00:01<00:31, 31.16it/s]\nSampling:   3%|▎         | 34/1000 [00:01<00:31, 31.09it/s]\nSampling:   4%|▍         | 38/1000 [00:01<00:30, 31.05it/s]\nSampling:   4%|▍         | 42/1000 [00:01<00:30, 31.60it/s]\nSampling:   5%|▍         | 46/1000 [00:01<00:30, 31.25it/s]\nSampling:   5%|▌         | 50/1000 [00:01<00:30, 31.61it/s]\nSampling:   5%|▌         | 54/1000 [00:01<00:29, 31.83it/s]\nSampling:   6%|▌         | 58/1000 [00:01<00:29, 32.00it/s]\nSampling:   6%|▌         | 62/1000 [00:02<00:29, 31.98it/s]\nSampling:   7%|▋         | 66/1000 [00:02<00:29, 32.10it/s]\nSampling:   7%|▋         | 70/1000 [00:02<00:29, 32.00it/s]\nSampling:   7%|▋         | 74/1000 [00:02<00:28, 32.18it/s]\nSampling:   8%|▊         | 78/1000 [00:02<00:29, 31.66it/s]\nSampling:   8%|▊         | 82/1000 [00:02<00:28, 31.85it/s]\nSampling:   9%|▊         | 86/1000 [00:02<00:28, 32.25it/s]\nSampling:   9%|▉         | 90/1000 [00:02<00:28, 32.31it/s]\nSampling:   9%|▉         | 94/1000 [00:03<00:28, 32.06it/s]\nSampling:  10%|▉         | 98/1000 [00:03<00:27, 32.33it/s]\nSampling:  10%|█         | 102/1000 [00:03<00:27, 32.26it/s]\nSampling:  11%|█         | 106/1000 [00:03<00:27, 32.38it/s]\nSampling:  11%|█         | 110/1000 [00:03<00:27, 32.14it/s]\nSampling:  11%|█▏        | 114/1000 [00:03<00:27, 32.41it/s]\nSampling:  12%|█▏        | 118/1000 [00:03<00:27, 32.64it/s]\nSampling:  12%|█▏        | 122/1000 [00:03<00:27, 32.15it/s]\nSampling:  13%|█▎        | 126/1000 [00:04<00:27, 32.24it/s]\nSampling:  13%|█▎        | 130/1000 [00:04<00:26, 32.24it/s]\nSampling:  13%|█▎        | 134/1000 [00:04<00:27, 31.22it/s]\nSampling:  14%|█▍        | 138/1000 [00:04<00:27, 31.70it/s]\nSampling:  14%|█▍        | 142/1000 [00:04<00:26, 31.80it/s]\nSampling:  15%|█▍        | 146/1000 [00:04<00:26, 32.17it/s]\nSampling:  15%|█▌        | 150/1000 [00:04<00:26, 31.71it/s]\nSampling:  15%|█▌        | 154/1000 [00:04<00:26, 31.60it/s]\nSampling:  16%|█▌        | 158/1000 [00:05<00:26, 31.53it/s]\nSampling:  16%|█▌        | 162/1000 [00:05<00:26, 31.55it/s]\nSampling:  17%|█▋        | 166/1000 [00:05<00:26, 31.52it/s]\nSampling:  17%|█▋        | 170/1000 [00:05<00:26, 31.89it/s]\nSampling:  17%|█▋        | 174/1000 [00:05<00:25, 31.97it/s]\nSampling:  18%|█▊        | 178/1000 [00:05<00:25, 32.06it/s]\nSampling:  18%|█▊        | 180/1000 [00:05<00:25, 31.54it/s]",
  "metrics": {
    "predict_time": 7.231277546,
    "total_time": 7.248154
  },
  "output": "https://replicate.delivery/czjl/WLb3N63OdupuNpN3owgmEAlLlAnWIPyTlt8ATeBw3hVCW7YKA/output.wav",
  "started_at": "2025-05-30T13:56:13.578877Z",
  "status": "succeeded",
  "urls": {
    "stream": "https://stream.replicate.com/v1/files/fddq-7tykk7bdcpig2qy7u5eyqqojrvuvhpu22kzmz65rek4wm7rcardq",
    "get": "https://api.replicate.com/v1/predictions/k0jzgr4yf9rga0cq45x9w5zfn8",
    "cancel": "https://api.replicate.com/v1/predictions/k0jzgr4yf9rga0cq45x9w5zfn8/cancel"
  },
  "version": "3f5f9c195086737dda710bf504330f71e786d0a361b505e377c8b10122af9d32"
}

Generated in

7.2 seconds

Tweak it ShareReport View full prediction

WARNING:root:Reference mel length is not equal to 2 * reference token length.
Sampling:   0%|          | 0/1000 [00:00<?, ?it/s]
Sampling:   0%|          | 2/1000 [00:00<00:58, 17.12it/s]
Sampling:   1%|          | 6/1000 [00:00<00:39, 25.40it/s]
Sampling:   1%|          | 10/1000 [00:00<00:34, 28.29it/s]
Sampling:   1%|▏         | 14/1000 [00:00<00:33, 29.14it/s]
Sampling:   2%|▏         | 18/1000 [00:00<00:33, 29.71it/s]
Sampling:   2%|▏         | 22/1000 [00:00<00:31, 30.60it/s]
Sampling:   3%|▎         | 26/1000 [00:00<00:31, 31.15it/s]
Sampling:   3%|▎         | 30/1000 [00:01<00:31, 31.16it/s]
Sampling:   3%|▎         | 34/1000 [00:01<00:31, 31.09it/s]
Sampling:   4%|▍         | 38/1000 [00:01<00:30, 31.05it/s]
Sampling:   4%|▍         | 42/1000 [00:01<00:30, 31.60it/s]
Sampling:   5%|▍         | 46/1000 [00:01<00:30, 31.25it/s]
Sampling:   5%|▌         | 50/1000 [00:01<00:30, 31.61it/s]
Sampling:   5%|▌         | 54/1000 [00:01<00:29, 31.83it/s]
Sampling:   6%|▌         | 58/1000 [00:01<00:29, 32.00it/s]
Sampling:   6%|▌         | 62/1000 [00:02<00:29, 31.98it/s]
Sampling:   7%|▋         | 66/1000 [00:02<00:29, 32.10it/s]
Sampling:   7%|▋         | 70/1000 [00:02<00:29, 32.00it/s]
Sampling:   7%|▋         | 74/1000 [00:02<00:28, 32.18it/s]
Sampling:   8%|▊         | 78/1000 [00:02<00:29, 31.66it/s]
Sampling:   8%|▊         | 82/1000 [00:02<00:28, 31.85it/s]
Sampling:   9%|▊         | 86/1000 [00:02<00:28, 32.25it/s]
Sampling:   9%|▉         | 90/1000 [00:02<00:28, 32.31it/s]
Sampling:   9%|▉         | 94/1000 [00:03<00:28, 32.06it/s]
Sampling:  10%|▉         | 98/1000 [00:03<00:27, 32.33it/s]
Sampling:  10%|█         | 102/1000 [00:03<00:27, 32.26it/s]
Sampling:  11%|█         | 106/1000 [00:03<00:27, 32.38it/s]
Sampling:  11%|█         | 110/1000 [00:03<00:27, 32.14it/s]
Sampling:  11%|█▏        | 114/1000 [00:03<00:27, 32.41it/s]
Sampling:  12%|█▏        | 118/1000 [00:03<00:27, 32.64it/s]
Sampling:  12%|█▏        | 122/1000 [00:03<00:27, 32.15it/s]
Sampling:  13%|█▎        | 126/1000 [00:04<00:27, 32.24it/s]
Sampling:  13%|█▎        | 130/1000 [00:04<00:26, 32.24it/s]
Sampling:  13%|█▎        | 134/1000 [00:04<00:27, 31.22it/s]
Sampling:  14%|█▍        | 138/1000 [00:04<00:27, 31.70it/s]
Sampling:  14%|█▍        | 142/1000 [00:04<00:26, 31.80it/s]
Sampling:  15%|█▍        | 146/1000 [00:04<00:26, 32.17it/s]
Sampling:  15%|█▌        | 150/1000 [00:04<00:26, 31.71it/s]
Sampling:  15%|█▌        | 154/1000 [00:04<00:26, 31.60it/s]
Sampling:  16%|█▌        | 158/1000 [00:05<00:26, 31.53it/s]
Sampling:  16%|█▌        | 162/1000 [00:05<00:26, 31.55it/s]
Sampling:  17%|█▋        | 166/1000 [00:05<00:26, 31.52it/s]
Sampling:  17%|█▋        | 170/1000 [00:05<00:26, 31.89it/s]
Sampling:  17%|█▋        | 174/1000 [00:05<00:25, 31.97it/s]
Sampling:  18%|█▊        | 178/1000 [00:05<00:25, 32.06it/s]
Sampling:  18%|█▊        | 180/1000 [00:05<00:25, 31.54it/s]

Run time and cost

This model costs approximately $0.029 to run on Replicate, or 34 runs per $1, but this varies depending on your inputs. It is also open source and you can run it on your own computer with Docker.

This model runs on Nvidia T4 GPU hardware. Predictions typically complete within 130 seconds. The predict time for this model varies significantly based on the inputs.

Readme

Chatterbox has been benchmarked against leading closed-source systems like ElevenLabs, and is consistently preferred in side-by-side evaluations.

Whether you’re working on memes, videos, games, or AI agents, Chatterbox brings your content to life. It’s also the first open source TTS model to support emotion exaggeration control, a powerful feature that makes your voices stand out. Try it now on our Hugging Face Gradio app.

Key Details

SoTA zeroshot TTS
0.5B Llama backbone
Unique exaggeration/intensity control
Ultra-stable with alignment-informed inference
Trained on 0.5M hours of cleaned data
Watermarked outputs
Easy voice conversion script
Outperforms ElevenLabs

Tips

General Use (TTS and Voice Agents):
The default settings (exaggeration=0.5, cfg_weight=0.5) work well for most prompts.
If the reference speaker has a fast speaking style, lowering cfg_weight to around 0.3 can improve pacing.
Expressive or Dramatic Speech:
Try lower cfg_weight values (e.g. ~0.3) and increase exaggeration to around 0.7 or higher.
Higher exaggeration tends to speed up speech; reducing cfg_weight helps compensate with slower, more deliberate pacing.

thomcle / chatterbox-tts

Input

Output

Run time and cost

Readme

Key Details

Tips

Logs (k0jzgr4yf9rga0cq45x9w5zfn8)