The Fish Speech V1.1 model.
Run this model in Node.js with one line of code:
npm install replicateREPLICATE_API_TOKEN environment variable:export REPLICATE_API_TOKEN=<paste-your-token-here>Find your API token in your account settings.
import Replicate from "replicate";
const replicate = new Replicate({
auth: process.env.REPLICATE_API_TOKEN,
});Run ttsds/fishspeech_1_1 using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.
const output = await replicate.run(
"ttsds/fishspeech_1_1:278768f6f5c2cc5a51d11dcf9fea9307063f6e053f489a9ccf16c46623e2a001",
{
input: {
text: "With tenure, Suzie'd have all the more leisure for yachting, but her publications are no good.",
text_reference: "and keeping eternity before the eyes, though much",
speaker_reference: "https://replicate.delivery/pbxt/MNFXdPaUPOwYCZjZM4azsymbzE2TCV2WJXfGpeV2DrFWaSq8/example_en.wav"
}
}
);
console.log(output);To learn more, take a look at the guide on getting started with Node.js.
pip install replicateREPLICATE_API_TOKEN environment variable:export REPLICATE_API_TOKEN=<paste-your-token-here>Find your API token in your account settings.
import replicateRun ttsds/fishspeech_1_1 using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.
output = replicate.run(
"ttsds/fishspeech_1_1:278768f6f5c2cc5a51d11dcf9fea9307063f6e053f489a9ccf16c46623e2a001",
input={
"text": "With tenure, Suzie'd have all the more leisure for yachting, but her publications are no good.",
"text_reference": "and keeping eternity before the eyes, though much",
"speaker_reference": "https://replicate.delivery/pbxt/MNFXdPaUPOwYCZjZM4azsymbzE2TCV2WJXfGpeV2DrFWaSq8/example_en.wav"
}
)
print(output)To learn more, take a look at the guide on getting started with Python.
REPLICATE_API_TOKEN environment variable:export REPLICATE_API_TOKEN=<paste-your-token-here>Find your API token in your account settings.
Run ttsds/fishspeech_1_1 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": "278768f6f5c2cc5a51d11dcf9fea9307063f6e053f489a9ccf16c46623e2a001",
"input": {
"text": "With tenure, Suzie\'d have all the more leisure for yachting, but her publications are no good.",
"text_reference": "and keeping eternity before the eyes, though much",
"speaker_reference": "https://replicate.delivery/pbxt/MNFXdPaUPOwYCZjZM4azsymbzE2TCV2WJXfGpeV2DrFWaSq8/example_en.wav"
}
}' \
https://api.replicate.com/v1/predictionsTo learn more, take a look at Replicate’s HTTP API reference docs.
brew install cogIf you don’t have Homebrew, there are other installation options available.
Run this to download the model and run it in your local environment:
cog predict r8.im/ttsds/fishspeech_1_1@sha256:278768f6f5c2cc5a51d11dcf9fea9307063f6e053f489a9ccf16c46623e2a001 \
-i $'text="With tenure, Suzie\'d have all the more leisure for yachting, but her publications are no good."' \
-i 'text_reference="and keeping eternity before the eyes, though much"' \
-i 'speaker_reference="https://replicate.delivery/pbxt/MNFXdPaUPOwYCZjZM4azsymbzE2TCV2WJXfGpeV2DrFWaSq8/example_en.wav"'To learn more, take a look at the Cog documentation.
Run this to download the model and run it in your local environment:
docker run -d -p 5000:5000 --gpus=all r8.im/ttsds/fishspeech_1_1@sha256:278768f6f5c2cc5a51d11dcf9fea9307063f6e053f489a9ccf16c46623e2a001curl -s -X POST \ -H "Content-Type: application/json" \ -d $'{ "input": { "text": "With tenure, Suzie\'d have all the more leisure for yachting, but her publications are no good.", "text_reference": "and keeping eternity before the eyes, though much", "speaker_reference": "https://replicate.delivery/pbxt/MNFXdPaUPOwYCZjZM4azsymbzE2TCV2WJXfGpeV2DrFWaSq8/example_en.wav" } }' \ http://localhost:5000/predictions
To learn more, take a look at the Cog documentation.
Add a payment method to run this model.
Each run costs approximately $0.0094. Alternatively, try out our featured models for free.
By signing in, you agree to our
terms of service and privacy policy
This is a modal window.
Beginning of dialog window. Escape will cancel and close the window.
End of dialog window.
{
"completed_at": "2025-01-28T16:26:05.034677Z",
"created_at": "2025-01-28T16:24:41.407000Z",
"data_removed": false,
"error": null,
"id": "fgdgr9djqxrme0cmnpx9bhp1em",
"input": {
"text": "With tenure, Suzie'd have all the more leisure for yachting, but her publications are no good.",
"text_reference": "and keeping eternity before the eyes, though much",
"speaker_reference": "https://replicate.delivery/pbxt/MNFXdPaUPOwYCZjZM4azsymbzE2TCV2WJXfGpeV2DrFWaSq8/example_en.wav"
},
"logs": "2025-01-28 16:26:02.222 | INFO | tools.llama.generate:generate_long:491 - Encoded text: With tenure, Suzie'd have all\n2025-01-28 16:26:02.222 | INFO | tools.llama.generate:generate_long:491 - Encoded text: the more leisure for yachting,\n2025-01-28 16:26:02.222 | INFO | tools.llama.generate:generate_long:491 - Encoded text: but her publications are no\n2025-01-28 16:26:02.222 | INFO | tools.llama.generate:generate_long:491 - Encoded text: good.\n2025-01-28 16:26:02.223 | INFO | tools.llama.generate:generate_long:509 - Generating sentence 1/4 of sample 1/1\n 0%| | 0/1858 [00:00<?, ?it/s]/root/.pyenv/versions/3.11.10/lib/python3.11/site-packages/torch/backends/cuda/__init__.py:342: FutureWarning: torch.backends.cuda.sdp_kernel() is deprecated. In the future, this context manager will be removed. Please see, torch.nn.attention.sdpa_kernel() for the new context manager, with updated signature.\nwarnings.warn(\n 0%| | 6/1858 [00:00<00:31, 59.00it/s]\n 1%| | 12/1858 [00:00<00:32, 57.47it/s]\n 1%| | 19/1858 [00:00<00:30, 60.07it/s]\n 1%|▏ | 26/1858 [00:00<00:29, 61.43it/s]\n 2%|▏ | 33/1858 [00:00<00:29, 62.05it/s]\n 2%|▏ | 40/1858 [00:00<00:29, 62.57it/s]\n 3%|▎ | 47/1858 [00:00<00:29, 62.38it/s]\n3%|▎ | 49/1858 [00:00<00:29, 60.40it/s]\n2025-01-28 16:26:03.118 | INFO | tools.llama.generate:generate_long:565 - Generated 51 tokens in 0.90 seconds, 56.95 tokens/sec\n2025-01-28 16:26:03.118 | INFO | tools.llama.generate:generate_long:568 - Bandwidth achieved: 22.22 GB/s\n2025-01-28 16:26:03.118 | INFO | tools.llama.generate:generate_long:573 - GPU Memory used: 2.45 GB\n2025-01-28 16:26:03.119 | INFO | tools.llama.generate:generate_long:509 - Generating sentence 2/4 of sample 1/1\n 0%| | 0/1759 [00:00<?, ?it/s]\n 0%| | 7/1759 [00:00<00:27, 62.93it/s]\n 1%| | 14/1759 [00:00<00:27, 62.96it/s]\n 1%| | 21/1759 [00:00<00:27, 62.82it/s]\n 2%|▏ | 28/1759 [00:00<00:28, 61.75it/s]\n 2%|▏ | 35/1759 [00:00<00:27, 62.10it/s]\n2%|▏ | 40/1759 [00:00<00:28, 60.85it/s]\n2025-01-28 16:26:03.792 | INFO | tools.llama.generate:generate_long:565 - Generated 42 tokens in 0.67 seconds, 62.37 tokens/sec\n2025-01-28 16:26:03.793 | INFO | tools.llama.generate:generate_long:568 - Bandwidth achieved: 24.34 GB/s\n2025-01-28 16:26:03.793 | INFO | tools.llama.generate:generate_long:573 - GPU Memory used: 2.45 GB\n2025-01-28 16:26:03.793 | INFO | tools.llama.generate:generate_long:509 - Generating sentence 3/4 of sample 1/1\n 0%| | 0/1672 [00:00<?, ?it/s]\n 0%| | 7/1672 [00:00<00:26, 63.08it/s]\n 1%| | 14/1672 [00:00<00:26, 62.91it/s]\n 1%|▏ | 21/1672 [00:00<00:26, 62.97it/s]\n 2%|▏ | 28/1672 [00:00<00:26, 62.73it/s]\n 2%|▏ | 35/1672 [00:00<00:26, 62.46it/s]\n2%|▏ | 39/1672 [00:00<00:26, 61.00it/s]\n2025-01-28 16:26:04.454 | INFO | tools.llama.generate:generate_long:565 - Generated 41 tokens in 0.66 seconds, 62.06 tokens/sec\n2025-01-28 16:26:04.454 | INFO | tools.llama.generate:generate_long:568 - Bandwidth achieved: 24.21 GB/s\n2025-01-28 16:26:04.454 | INFO | tools.llama.generate:generate_long:573 - GPU Memory used: 2.45 GB\n2025-01-28 16:26:04.454 | INFO | tools.llama.generate:generate_long:509 - Generating sentence 4/4 of sample 1/1\n 0%| | 0/1608 [00:00<?, ?it/s]\n 0%| | 7/1608 [00:00<00:25, 63.11it/s]\n 1%| | 14/1608 [00:00<00:25, 63.00it/s]\n1%| | 14/1608 [00:00<00:27, 58.74it/s]\n2025-01-28 16:26:04.709 | INFO | tools.llama.generate:generate_long:565 - Generated 16 tokens in 0.25 seconds, 62.81 tokens/sec\n2025-01-28 16:26:04.709 | INFO | tools.llama.generate:generate_long:568 - Bandwidth achieved: 24.51 GB/s\n2025-01-28 16:26:04.710 | INFO | tools.llama.generate:generate_long:573 - GPU Memory used: 2.45 GB\n/root/.pyenv/versions/3.11.10/lib/python3.11/site-packages/torch/nn/modules/conv.py:306: UserWarning: Plan failed with a cudnnException: CUDNN_BACKEND_EXECUTION_PLAN_DESCRIPTOR: cudnnFinalize Descriptor Failed cudnn_status: CUDNN_STATUS_NOT_SUPPORTED (Triggered internally at ../aten/src/ATen/native/cudnn/Conv_v8.cpp:919.)\nreturn F.conv1d(input, weight, bias, self.stride,\nNext sample",
"metrics": {
"predict_time": 3.097940238,
"total_time": 83.627677
},
"output": "https://replicate.delivery/xezq/I77Wgm5GPZJcPBiGe9WeRYHDfLr5blb16hWlvxnkk3f3xtmQB/generated.wav",
"started_at": "2025-01-28T16:26:01.936737Z",
"status": "succeeded",
"urls": {
"stream": "https://stream.replicate.com/v1/files/bsvm-kkjxzcufksibrwqp56dqg3nxd4bsxbkgeiscttvpaqnc4nagf4wa",
"get": "https://api.replicate.com/v1/predictions/fgdgr9djqxrme0cmnpx9bhp1em",
"cancel": "https://api.replicate.com/v1/predictions/fgdgr9djqxrme0cmnpx9bhp1em/cancel"
},
"version": "eba2a3e1e07cf38ac2a528d134fdac1cab4d222b8850ac98517635ddf5c1ca75"
}2025-01-28 16:26:02.222 | INFO | tools.llama.generate:generate_long:491 - Encoded text: With tenure, Suzie'd have all
2025-01-28 16:26:02.222 | INFO | tools.llama.generate:generate_long:491 - Encoded text: the more leisure for yachting,
2025-01-28 16:26:02.222 | INFO | tools.llama.generate:generate_long:491 - Encoded text: but her publications are no
2025-01-28 16:26:02.222 | INFO | tools.llama.generate:generate_long:491 - Encoded text: good.
2025-01-28 16:26:02.223 | INFO | tools.llama.generate:generate_long:509 - Generating sentence 1/4 of sample 1/1
0%| | 0/1858 [00:00<?, ?it/s]/root/.pyenv/versions/3.11.10/lib/python3.11/site-packages/torch/backends/cuda/__init__.py:342: FutureWarning: torch.backends.cuda.sdp_kernel() is deprecated. In the future, this context manager will be removed. Please see, torch.nn.attention.sdpa_kernel() for the new context manager, with updated signature.
warnings.warn(
0%| | 6/1858 [00:00<00:31, 59.00it/s]
1%| | 12/1858 [00:00<00:32, 57.47it/s]
1%| | 19/1858 [00:00<00:30, 60.07it/s]
1%|▏ | 26/1858 [00:00<00:29, 61.43it/s]
2%|▏ | 33/1858 [00:00<00:29, 62.05it/s]
2%|▏ | 40/1858 [00:00<00:29, 62.57it/s]
3%|▎ | 47/1858 [00:00<00:29, 62.38it/s]
3%|▎ | 49/1858 [00:00<00:29, 60.40it/s]
2025-01-28 16:26:03.118 | INFO | tools.llama.generate:generate_long:565 - Generated 51 tokens in 0.90 seconds, 56.95 tokens/sec
2025-01-28 16:26:03.118 | INFO | tools.llama.generate:generate_long:568 - Bandwidth achieved: 22.22 GB/s
2025-01-28 16:26:03.118 | INFO | tools.llama.generate:generate_long:573 - GPU Memory used: 2.45 GB
2025-01-28 16:26:03.119 | INFO | tools.llama.generate:generate_long:509 - Generating sentence 2/4 of sample 1/1
0%| | 0/1759 [00:00<?, ?it/s]
0%| | 7/1759 [00:00<00:27, 62.93it/s]
1%| | 14/1759 [00:00<00:27, 62.96it/s]
1%| | 21/1759 [00:00<00:27, 62.82it/s]
2%|▏ | 28/1759 [00:00<00:28, 61.75it/s]
2%|▏ | 35/1759 [00:00<00:27, 62.10it/s]
2%|▏ | 40/1759 [00:00<00:28, 60.85it/s]
2025-01-28 16:26:03.792 | INFO | tools.llama.generate:generate_long:565 - Generated 42 tokens in 0.67 seconds, 62.37 tokens/sec
2025-01-28 16:26:03.793 | INFO | tools.llama.generate:generate_long:568 - Bandwidth achieved: 24.34 GB/s
2025-01-28 16:26:03.793 | INFO | tools.llama.generate:generate_long:573 - GPU Memory used: 2.45 GB
2025-01-28 16:26:03.793 | INFO | tools.llama.generate:generate_long:509 - Generating sentence 3/4 of sample 1/1
0%| | 0/1672 [00:00<?, ?it/s]
0%| | 7/1672 [00:00<00:26, 63.08it/s]
1%| | 14/1672 [00:00<00:26, 62.91it/s]
1%|▏ | 21/1672 [00:00<00:26, 62.97it/s]
2%|▏ | 28/1672 [00:00<00:26, 62.73it/s]
2%|▏ | 35/1672 [00:00<00:26, 62.46it/s]
2%|▏ | 39/1672 [00:00<00:26, 61.00it/s]
2025-01-28 16:26:04.454 | INFO | tools.llama.generate:generate_long:565 - Generated 41 tokens in 0.66 seconds, 62.06 tokens/sec
2025-01-28 16:26:04.454 | INFO | tools.llama.generate:generate_long:568 - Bandwidth achieved: 24.21 GB/s
2025-01-28 16:26:04.454 | INFO | tools.llama.generate:generate_long:573 - GPU Memory used: 2.45 GB
2025-01-28 16:26:04.454 | INFO | tools.llama.generate:generate_long:509 - Generating sentence 4/4 of sample 1/1
0%| | 0/1608 [00:00<?, ?it/s]
0%| | 7/1608 [00:00<00:25, 63.11it/s]
1%| | 14/1608 [00:00<00:25, 63.00it/s]
1%| | 14/1608 [00:00<00:27, 58.74it/s]
2025-01-28 16:26:04.709 | INFO | tools.llama.generate:generate_long:565 - Generated 16 tokens in 0.25 seconds, 62.81 tokens/sec
2025-01-28 16:26:04.709 | INFO | tools.llama.generate:generate_long:568 - Bandwidth achieved: 24.51 GB/s
2025-01-28 16:26:04.710 | INFO | tools.llama.generate:generate_long:573 - GPU Memory used: 2.45 GB
/root/.pyenv/versions/3.11.10/lib/python3.11/site-packages/torch/nn/modules/conv.py:306: UserWarning: Plan failed with a cudnnException: CUDNN_BACKEND_EXECUTION_PLAN_DESCRIPTOR: cudnnFinalize Descriptor Failed cudnn_status: CUDNN_STATUS_NOT_SUPPORTED (Triggered internally at ../aten/src/ATen/native/cudnn/Conv_v8.cpp:919.)
return F.conv1d(input, weight, bias, self.stride,
Next sampleThis example was created by a different version, ttsds/fishspeech_1_1:eba2a3e1.
This model costs approximately $0.0094 to run on Replicate, or 106 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 A100 (80GB) GPU hardware. Predictions typically complete within 7 seconds.
The 1.1 version of the model released by fish.audio.
This is an unofficial implementation. Please refer to the fishspeech repository for the original code and details. By using this model, you agree to the terms stated at the link above, which could change at any time. Make sure to comply to these terms before using the model.
This model is cold. You'll get a fast response if the model is warm and already running, and a slower response if the model is cold and starting up.
Choose a file from your machine
Hint: you can also drag files onto the input
2025-01-28 16:26:02.222 | INFO | tools.llama.generate:generate_long:491 - Encoded text: With tenure, Suzie'd have all
2025-01-28 16:26:02.222 | INFO | tools.llama.generate:generate_long:491 - Encoded text: the more leisure for yachting,
2025-01-28 16:26:02.222 | INFO | tools.llama.generate:generate_long:491 - Encoded text: but her publications are no
2025-01-28 16:26:02.222 | INFO | tools.llama.generate:generate_long:491 - Encoded text: good.
2025-01-28 16:26:02.223 | INFO | tools.llama.generate:generate_long:509 - Generating sentence 1/4 of sample 1/1
0%| | 0/1858 [00:00<?, ?it/s]/root/.pyenv/versions/3.11.10/lib/python3.11/site-packages/torch/backends/cuda/__init__.py:342: FutureWarning: torch.backends.cuda.sdp_kernel() is deprecated. In the future, this context manager will be removed. Please see, torch.nn.attention.sdpa_kernel() for the new context manager, with updated signature.
warnings.warn(
0%| | 6/1858 [00:00<00:31, 59.00it/s]
1%| | 12/1858 [00:00<00:32, 57.47it/s]
1%| | 19/1858 [00:00<00:30, 60.07it/s]
1%|▏ | 26/1858 [00:00<00:29, 61.43it/s]
2%|▏ | 33/1858 [00:00<00:29, 62.05it/s]
2%|▏ | 40/1858 [00:00<00:29, 62.57it/s]
3%|▎ | 47/1858 [00:00<00:29, 62.38it/s]
3%|▎ | 49/1858 [00:00<00:29, 60.40it/s]
2025-01-28 16:26:03.118 | INFO | tools.llama.generate:generate_long:565 - Generated 51 tokens in 0.90 seconds, 56.95 tokens/sec
2025-01-28 16:26:03.118 | INFO | tools.llama.generate:generate_long:568 - Bandwidth achieved: 22.22 GB/s
2025-01-28 16:26:03.118 | INFO | tools.llama.generate:generate_long:573 - GPU Memory used: 2.45 GB
2025-01-28 16:26:03.119 | INFO | tools.llama.generate:generate_long:509 - Generating sentence 2/4 of sample 1/1
0%| | 0/1759 [00:00<?, ?it/s]
0%| | 7/1759 [00:00<00:27, 62.93it/s]
1%| | 14/1759 [00:00<00:27, 62.96it/s]
1%| | 21/1759 [00:00<00:27, 62.82it/s]
2%|▏ | 28/1759 [00:00<00:28, 61.75it/s]
2%|▏ | 35/1759 [00:00<00:27, 62.10it/s]
2%|▏ | 40/1759 [00:00<00:28, 60.85it/s]
2025-01-28 16:26:03.792 | INFO | tools.llama.generate:generate_long:565 - Generated 42 tokens in 0.67 seconds, 62.37 tokens/sec
2025-01-28 16:26:03.793 | INFO | tools.llama.generate:generate_long:568 - Bandwidth achieved: 24.34 GB/s
2025-01-28 16:26:03.793 | INFO | tools.llama.generate:generate_long:573 - GPU Memory used: 2.45 GB
2025-01-28 16:26:03.793 | INFO | tools.llama.generate:generate_long:509 - Generating sentence 3/4 of sample 1/1
0%| | 0/1672 [00:00<?, ?it/s]
0%| | 7/1672 [00:00<00:26, 63.08it/s]
1%| | 14/1672 [00:00<00:26, 62.91it/s]
1%|▏ | 21/1672 [00:00<00:26, 62.97it/s]
2%|▏ | 28/1672 [00:00<00:26, 62.73it/s]
2%|▏ | 35/1672 [00:00<00:26, 62.46it/s]
2%|▏ | 39/1672 [00:00<00:26, 61.00it/s]
2025-01-28 16:26:04.454 | INFO | tools.llama.generate:generate_long:565 - Generated 41 tokens in 0.66 seconds, 62.06 tokens/sec
2025-01-28 16:26:04.454 | INFO | tools.llama.generate:generate_long:568 - Bandwidth achieved: 24.21 GB/s
2025-01-28 16:26:04.454 | INFO | tools.llama.generate:generate_long:573 - GPU Memory used: 2.45 GB
2025-01-28 16:26:04.454 | INFO | tools.llama.generate:generate_long:509 - Generating sentence 4/4 of sample 1/1
0%| | 0/1608 [00:00<?, ?it/s]
0%| | 7/1608 [00:00<00:25, 63.11it/s]
1%| | 14/1608 [00:00<00:25, 63.00it/s]
1%| | 14/1608 [00:00<00:27, 58.74it/s]
2025-01-28 16:26:04.709 | INFO | tools.llama.generate:generate_long:565 - Generated 16 tokens in 0.25 seconds, 62.81 tokens/sec
2025-01-28 16:26:04.709 | INFO | tools.llama.generate:generate_long:568 - Bandwidth achieved: 24.51 GB/s
2025-01-28 16:26:04.710 | INFO | tools.llama.generate:generate_long:573 - GPU Memory used: 2.45 GB
/root/.pyenv/versions/3.11.10/lib/python3.11/site-packages/torch/nn/modules/conv.py:306: UserWarning: Plan failed with a cudnnException: CUDNN_BACKEND_EXECUTION_PLAN_DESCRIPTOR: cudnnFinalize Descriptor Failed cudnn_status: CUDNN_STATUS_NOT_SUPPORTED (Triggered internally at ../aten/src/ATen/native/cudnn/Conv_v8.cpp:919.)
return F.conv1d(input, weight, bias, self.stride,
Next sample