Fish-Speech 1.4

Open weights Fish Audio November 2024

No estimate

No hardware requirements for this model

This model's weights are open, but no parameter count has been published for it. Every memory and speed figure starts from that number, so we would rather show nothing than a fabricated estimate.

On record

Full specification

Everything on record for this model. Most of it describes how it was trained rather than how it runs — useful context for judging how much work went into it, and how it compares with models built at a different scale.

Origin

Who built this model, where, and when it was published.

Organisation
Fish Audio
Organisation type
Industry
Country
United States of America
Published
9 November 2024
Authors
Shijia Liao, Yuxuan Wang, Tianyu Li, Yifan Cheng, Ruoyi Zhang, Rongzhi Zhou, Yijin Xing

What it does

The problem areas the model was built for. A model can carry several of each.

Domain
Speech
Task
Speech synthesis, Text-to-speech (TTS)

Size

How large the model is and how much data it was trained on. Parameters are the figure that decides whether it fits on a given graphics card.

Training data
500,000,000,000 tokens

[tokens] "The dataset contains about 720,000 hours of speech across different languages, with 300,000 hours each of English and Mandarin Chinese as the main components. We also included 20,000 hours each of other language families: Germanic (German), Romance (French, Italian), East Asian (Japanese, Korean), and Semitic (Arabic)." • Batch size: 1M tokens • Training steps: 500K 10^6 * 500000 = 5*10^11 tokens

Training compute

The arithmetic performed to train the model, measured in floating-point operations. It is a measure of what the training run cost, not of how fast the finished model answers you.

Training compute
1.9 × 10²¹ FLOP

"The AR training utilized an 8*H100 80G GPUs for one week, while the vocoder training employed an 8*4090 GPUs for an additional week. Note that these timelines exclude the DPO stage." (989400000000000 FLOP / GPU / sec [H100, bf16 assumed] + 330000000000000 FLOP / GPU / sec [4090, bf16 assumed]) * 8 GPUs * 168 hours * 3600 sec / hour * 0.3 [assumed utilization] = 1.9151355e+21 FLOP

How it was established
Hardware

The training run

What it physically took to train: which chips, how many, for how long, and what that drew from the wall.

Training hardware
NVIDIA H100 SXM5 80GB,NVIDIA GeForce RTX 4090
Chips used
8
Wall-clock time
168 hours (7 days)

"The AR training utilized an 8*H100 80G GPUs for one week, while the vocoder training employed an 8*4090 GPUs for an additional week" 1 week = 168 hours

Availability

Whether you can obtain the model and run it on your own hardware, which is what decides if any of the graphics-card figures on this page apply.

Weights
Open — downloadable
Model access
Open weights (non-commercial)
Training code
Open source

This codebase is released under Apache License and all model weights are released under CC-BY-NC-SA-4.0 License https://github.com/fishaudio/fish-speech/tree/main/fish_speech https://huggingface.co/fishaudio/fish-speech-1.4

Hugging Face
fishaudio

How it is classified

Labels the source dataset applies when tracking notable models, and how confident it is in the entry.

Record confidence
Confident

Sources

Where this record came from and when it was last checked.

Reference
Fish-Speech: Leveraging Large Language Models for Advanced Multilingual Text-to-Speech Synthesis
Last updated
28 November 2025

What the numbers mean

Where it came from

Fish-Speech 1.4 was published by Fish Audio, in United States of America, in November 2024. The organisation is categorised as industry.

It works in Speech, and is recorded as doing speech synthesis, Text-to-speech (TTS).

Because its weights were released, nothing about running it depends on a provider staying available — it is yours once downloaded. It is published under the fishaudio organisation on Hugging Face.

What went into building it

Training it took roughly 1.9 × 10²¹ FLOP of computation, on NVIDIA H100 SXM5 80GB,NVIDIA GeForce RTX 4090 — a measure of what producing the model cost, not of how fast it answers.

It was trained on about 500,000,000,000 tokens of text.

Answers

Fish-Speech 1.4 — common questions

01

When was Fish-Speech 1.4 released?

Fish-Speech 1.4 was published in November 2024. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

02

What is Fish-Speech 1.4 used for?

Fish-Speech 1.4 works in Speech, and is recorded as handling speech synthesis, Text-to-speech (TTS). A model can carry several of each, so these are the areas it was built for rather than a limit on what it will attempt.

03

Where can I download Fish-Speech 1.4?

Its weights are published under the fishaudio organisation on Hugging Face. We do not host model files — this site calculates what hardware is needed to run them.

04

How much compute was used to train Fish-Speech 1.4?

Around 1.9 × 10²¹ FLOP, on NVIDIA H100 SXM5 80GB,NVIDIA GeForce RTX 4090. That measures what producing the model cost and says nothing about how quickly it answers once trained — inference speed comes from memory bandwidth, not from the training budget.

05

What GPU do I need to run Fish-Speech 1.4?

We cannot say. Fish-Speech 1.4 has open weights, but no parameter count has been published for it, and every memory and speed calculation starts from that number. We would rather show nothing than a fabricated estimate.

06

Is Fish-Speech 1.4 open source?

Its weights are published, so Fish-Speech 1.4 can be downloaded and run on your own hardware. Note that open weights is not the same as open source in the full sense — it says nothing about the training data, the training code, or the commercial terms attached.

07

How many parameters does Fish-Speech 1.4 have?

No parameter count has been published for Fish-Speech 1.4, which is why no memory or speed figure appears on this page.

08

Who created Fish-Speech 1.4?

Fish-Speech 1.4 was published by Fish Audio, based in United States of America, categorised as industry.

Source

Original publication

Record last updated 28 November 2025

The other direction

Looking at it from the other side?

This page starts from the model. If you already own a card and want to know everything it will run, start from the hardware instead.