indic-tts-mini-train by Sulabh Katiyar: Open-Weight Model

Model Card

By Sulabh Katiyar, published under cc-by-4.0, revision 009a1448a6c7.

This repository holds intermediate training checkpoints, logs and metrics of the Mini TTS v1 research pilot: a small multilingual Matcha-style acoustic model (MAS + OT-CFM, 9M and 26M parameters) for Indian languages and English. The checkpoints are experimental, may be deleted or overwritten at any time, and are not intended for use. There is no quality claim and no support. - Rasa (AI4Bharat), https://huggingface.co/datasets/ai4bharat/Rasa, licensed CC-BY-4.0. Changes: the audio was resampled from 48 kHz to 24 kHz, emotional styles were filtered out, and each voice was capped (pilot: 8 h per voice). The derived dataset is sulabhkatiyar/indic-tts-mini-24k. - IndicTTS-English (pilot only…

Read Sulabh Katiyar's full model card

indic-tts-mini-train: research checkpoints, NOT a release

This repository holds intermediate training checkpoints, logs and metrics of the Mini TTS v1 research pilot: a small multilingual Matcha-style acoustic model (MAS + OT-CFM, 9M and 26M parameters) for Indian languages and English. The checkpoints are experimental, may be deleted or overwritten at any time, and are not intended for use. There is no quality claim and no support.

Training data and licences

Rasa (AI4Bharat), https://huggingface.co/datasets/ai4bharat/Rasa, licensed CC-BY-4.0. Changes: the audio was resampled from 48 kHz to 24 kHz, emotional styles were filtered out, and each voice was capped (pilot: 8 h per voice). The derived dataset is sulabhkatiyar/indic-tts-mini-24k.
IndicTTS-English (pilot only, never released), Indic TTS database, IIT Madras: "COPYRIGHT 2016 TTS Consortium, TDIL, Meity", IIT Madras (see the IndicTTS database licence agreement for the full notice). It is used here only in the research pilot. The final models will instead use LIMMITS/SYSPIN English (IISc, CC-BY-4.0).
English phone sequences come from eSpeak NG (GPL-3.0). It is used offline as a data tool only.

Code and models

Acoustic model code is derived from Matcha-TTS (https://github.com/shivammehta25/Matcha-TTS, MIT licence, Copyright (c) 2023 Shivam Mehta).
The vocoder used for listening checks is Vocos (https://github.com/gemelo-ai/vocos, MIT licence, Copyright (c) 2023 Charactr Inc.) with the pretrained charactr/vocos-mel-24khz weights (MIT). It is trained on LibriTTS, which is CC-BY-4.0.

Layout

runs/<run>/ckpt_*.pt and latest.json: the newest rolling checkpoints (full state, used for resume).
runs/<run>/milestones/: permanent weight snapshots (model + EMA), taken at fixed step intervals.
runs/<run>/train_log.jsonl / metrics.jsonl: training and validation metrics.

Identity and Version

Repository: sulabhkatiyar/indic-tts-mini-train
Publisher: Sulabh Katiyar
Task: Text to speech
Modality: Audio
Library: Not stated by the source
Parameters: Not stated by the source
Languages: Not stated by the source
Revision: 009a1448a6c7465c0c37dd333e49a1064f3eeccb
First published: 2026-10-15
Last updated: 2026-10-16

Files and Weights

95 files, 8.6 GB in total. The weights are 79 files totalling 8.6 GB in pt.

Weights79 files · 8.6 GB

Configuration8 files · 1.6 KB

Documentation1 file · 2.0 KB

Other6 files · 2.4 MB

Repository1 file · 1.5 KB

Every file

File	Type	Size	SHA-256
runs/A-chr-P1a-v2/ckpt_00080592.pt	Weights	143.8 MB	ff30da8593dc
runs/A-chr-P1a-v2/ckpt_00086860.pt	Weights	143.8 MB	c543b1a31c27
runs/A-chr-P1a-v2/milestones/ms_00005000.pt	Weights	71.9 MB	4505cbe2d602
runs/A-chr-P1a-v2/milestones/ms_00010000.pt	Weights	71.9 MB	ada33ded364d
runs/A-chr-P1a-v2/milestones/ms_00015000.pt	Weights	71.9 MB	f143b96d374e
runs/A-chr-P1a-v2/milestones/ms_00020000.pt	Weights	71.9 MB	851fa8d389da
runs/A-chr-P1a-v2/milestones/ms_00025000.pt	Weights	71.9 MB	7ff3b3e02da6
runs/A-chr-P1a-v2/milestones/ms_00030000.pt	Weights	71.9 MB	1fd8edaf2c72
runs/A-chr-P1a-v2/milestones/ms_00035000.pt	Weights	71.9 MB	b669f6f8ea93
runs/A-chr-P1a-v2/milestones/ms_00040000.pt	Weights	71.9 MB	087811e2bf16
runs/A-chr-P1a-v2/milestones/ms_00045000.pt	Weights	71.9 MB	53c2c8fa4cae
runs/A-chr-P1a-v2/milestones/ms_00050000.pt	Weights	71.9 MB	ccdfe8be7605
runs/A-chr-P1a-v2/milestones/ms_00055000.pt	Weights	71.9 MB	892985dbd2bb
runs/A-chr-P1a-v2/milestones/ms_00060000.pt	Weights	71.9 MB	61df6bdccbe4
runs/A-chr-P1a-v2/milestones/ms_00065000.pt	Weights	71.9 MB	3f7444842859
runs/A-chr-P1a-v2/milestones/ms_00070000.pt	Weights	71.9 MB	09c5c76a1d56
runs/A-chr-P1a-v2/milestones/ms_00075000.pt	Weights	71.9 MB	6e08eaeea174
runs/A-chr-P1a-v2/milestones/ms_00080000.pt	Weights	71.9 MB	54360bbdd3af
runs/A-chr-P1a-v2/milestones/ms_00085000.pt	Weights	71.9 MB	aae3d38cf7b3
runs/A-ipa-P1a-v2/ckpt_00082209.pt	Weights	143.5 MB	e023f1cf2ac8
runs/A-ipa-P1a-v2/ckpt_00088601.pt	Weights	143.5 MB	c048fd09a904
runs/A-ipa-P1a-v2/milestones/ms_00005000.pt	Weights	71.7 MB	c80ff8bcdfac
runs/A-ipa-P1a-v2/milestones/ms_00010000.pt	Weights	71.7 MB	4473c1c9682a
runs/A-ipa-P1a-v2/milestones/ms_00015000.pt	Weights	71.7 MB	3888b57d7b4e
runs/A-ipa-P1a-v2/milestones/ms_00020000.pt	Weights	71.7 MB	178b7dff2ab3
runs/A-ipa-P1a-v2/milestones/ms_00025000.pt	Weights	71.7 MB	24ee0e07cd17
runs/A-ipa-P1a-v2/milestones/ms_00030000.pt	Weights	71.7 MB	b3b24a3d09c3
runs/A-ipa-P1a-v2/milestones/ms_00035000.pt	Weights	71.7 MB	cb2555ff8fad
runs/A-ipa-P1a-v2/milestones/ms_00040000.pt	Weights	71.7 MB	97159b6e4ad2
runs/A-ipa-P1a-v2/milestones/ms_00045000.pt	Weights	71.7 MB	71262470fbb1
runs/A-ipa-P1a-v2/milestones/ms_00050000.pt	Weights	71.7 MB	2a108df79aa7
runs/A-ipa-P1a-v2/milestones/ms_00055000.pt	Weights	71.7 MB	5b6ffd7de4c7
runs/A-ipa-P1a-v2/milestones/ms_00060000.pt	Weights	71.7 MB	2b9c0732be51
runs/A-ipa-P1a-v2/milestones/ms_00065000.pt	Weights	71.7 MB	7a3b3ab459a6
runs/A-ipa-P1a-v2/milestones/ms_00070000.pt	Weights	71.7 MB	b10bf4a6f5c5
runs/A-ipa-P1a-v2/milestones/ms_00075000.pt	Weights	71.7 MB	d44f501ba431
runs/A-ipa-P1a-v2/milestones/ms_00080000.pt	Weights	71.7 MB	fde0eee1487d
runs/A-ipa-P1a-v2/milestones/ms_00085000.pt	Weights	71.7 MB	722aaec5983a
runs/B-hi-P1b-v2/ckpt_00084744.pt	Weights	143.4 MB	d08aefa6c23e
runs/B-hi-P1b-v2/ckpt_00091795.pt	Weights	143.4 MB	f7670dfffe1d
runs/B-hi-P1b-v2/milestones/ms_00005000.pt	Weights	71.7 MB	216280fb2920
runs/B-hi-P1b-v2/milestones/ms_00010000.pt	Weights	71.7 MB	9d388d5fa6b1
runs/B-hi-P1b-v2/milestones/ms_00015000.pt	Weights	71.7 MB	3f87b787a227
runs/B-hi-P1b-v2/milestones/ms_00020000.pt	Weights	71.7 MB	f61e8f2854bf
runs/B-hi-P1b-v2/milestones/ms_00025000.pt	Weights	71.7 MB	b9fb705af42c
runs/B-hi-P1b-v2/milestones/ms_00030000.pt	Weights	71.7 MB	97b4b80fdf34
runs/B-hi-P1b-v2/milestones/ms_00035000.pt	Weights	71.7 MB	adec45b27f9f
runs/B-hi-P1b-v2/milestones/ms_00040000.pt	Weights	71.7 MB	05381c97b759
runs/B-hi-P1b-v2/milestones/ms_00045000.pt	Weights	71.7 MB	7b3006d48552
runs/B-hi-P1b-v2/milestones/ms_00050000.pt	Weights	71.7 MB	384424f4b008
runs/B-hi-P1b-v2/milestones/ms_00055000.pt	Weights	71.7 MB	3eda57664dc9
runs/B-hi-P1b-v2/milestones/ms_00060000.pt	Weights	71.7 MB	e93f33183c8c
runs/B-hi-P1b-v2/milestones/ms_00065000.pt	Weights	71.7 MB	6530c7812f8c
runs/B-hi-P1b-v2/milestones/ms_00070000.pt	Weights	71.7 MB	6dbf9815802b
runs/B-hi-P1b-v2/milestones/ms_00075000.pt	Weights	71.7 MB	5b84e5e06fc8
runs/B-hi-P1b-v2/milestones/ms_00080000.pt	Weights	71.7 MB	73a8ed7967c1
runs/B-hi-P1b-v2/milestones/ms_00085000.pt	Weights	71.7 MB	b7aa342bf606
runs/B-hi-P1b-v2/milestones/ms_00090000.pt	Weights	71.7 MB	2082a6957fdf
runs/W-8-P1b-v2/ckpt_00055223.pt	Weights	419.3 MB	14d6a8f4a795
runs/W-8-P1b-v2/ckpt_00059846.pt	Weights	419.3 MB	c686c5051489
runs/W-8-P1b-v2/milestones/ms_00005000.pt	Weights	209.6 MB	b2f03db6b7b6
runs/W-8-P1b-v2/milestones/ms_00010000.pt	Weights	209.6 MB	4cf237fd82db
runs/W-8-P1b-v2/milestones/ms_00015000.pt	Weights	209.6 MB	b358a714bd08
runs/W-8-P1b-v2/milestones/ms_00020000.pt	Weights	209.6 MB	b47ae142c223
runs/W-8-P1b-v2/milestones/ms_00025000.pt	Weights	209.6 MB	a523cdd25605
runs/W-8-P1b-v2/milestones/ms_00030000.pt	Weights	209.6 MB	5d5734f3f5b0
runs/W-8-P1b-v2/milestones/ms_00035000.pt	Weights	209.6 MB	d99e684bfab4
runs/W-8-P1b-v2/milestones/ms_00040000.pt	Weights	209.6 MB	dcae21fed227
runs/W-8-P1b-v2/milestones/ms_00045000.pt	Weights	209.6 MB	895ca60b68ac
runs/W-8-P1b-v2/milestones/ms_00050000.pt	Weights	209.6 MB	bfc7b927ba54
runs/W-8-P1b-v2/milestones/ms_00055000.pt	Weights	209.6 MB	728661785d0e
runs/probe-resume-A-chr-1792098895/ckpt_00000200.pt	Weights	143.8 MB	d85622b3deaf
runs/probe-resume-A-chr-1792098895/ckpt_00000400.pt	Weights	143.8 MB	8d23d2b3f08e
runs/probe-resume-A-chr-1792098895/milestones/ms_00000200.pt	Weights	71.9 MB	90cabdeb068e
runs/probe-resume-A-chr-1792098895/milestones/ms_00000400.pt	Weights	71.9 MB	87056731eb34
runs/probe-resume-A-ipa-1792098895/ckpt_00000200.pt	Weights	143.5 MB	d04254a40455
runs/probe-resume-A-ipa-1792098895/ckpt_00000400.pt	Weights	143.5 MB	b3d0378d73b5
runs/probe-resume-A-ipa-1792098895/milestones/ms_00000200.pt	Weights	71.7 MB	5b48768294fc
runs/probe-resume-A-ipa-1792098895/milestones/ms_00000400.pt	Weights	71.7 MB	ad1280ed582f
runs/A-chr-P1a-v2/latest.json	Configuration	155 B	—
runs/A-ipa-P1a-v2/latest.json	Configuration	155 B	—
runs/B-hi-P1b-v2/latest.json	Configuration	155 B	—
runs/W-8-P1b-v2/latest.json	Configuration	155 B	—
runs/probe-resume-A-chr-1792098895/DONE.json	Configuration	348 B	—
runs/probe-resume-A-chr-1792098895/latest.json	Configuration	153 B	—
runs/probe-resume-A-ipa-1792098895/DONE.json	Configuration	349 B	—
runs/probe-resume-A-ipa-1792098895/latest.json	Configuration	153 B	—
README.md	Documentation	2.0 KB	—
runs/A-chr-P1a-v2/train_log.jsonl	Other	633.5 KB	—
runs/A-ipa-P1a-v2/train_log.jsonl	Other	646.3 KB	—
runs/B-hi-P1b-v2/train_log.jsonl	Other	670.0 KB	—
runs/W-8-P1b-v2/train_log.jsonl	Other	435.2 KB	—
runs/probe-resume-A-chr-1792098895/train_log.jsonl	Other	2.9 KB	—
runs/probe-resume-A-ipa-1792098895/train_log.jsonl	Other	2.9 KB	—
.gitattributes	Repository	1.5 KB	—

License and Download

License: cc-by-4.0
Access: Open weights, no gate
Download size: 8.6 GB

Download from Sulabh Katiyar

Released by Sulabh Katiyar through its official repository on Hugging Face. Read the license.

Memory Requirements

Precision	Weights in memory
As published	8.6 GB

Weights only, from the published parameter count; the key-value cache and runtime add to this.

Questions About indic-tts-mini-train

Can I use indic-tts-mini-train commercially?

Yes. indic-tts-mini-train is released under Creative Commons Attribution 4.0. CC BY 4.0 permits sharing and adapting the work, including commercially, provided the creator is credited and changes are indicated.

Similar Models

Model · Text to speech

Kokoro-82M

Hexgrad

Kokoro is an open-weight TTS model with 82 million parameters. Despite its lightweight architecture, it delivers comparable quality to larger models while being significantly faster and more cost-efficient. With Apache-licensed weights, Kokoro can be deployed anywhere from production environments to personal projects. You can run this basic cell on Google Colab. Listen to samples. For more languages and details, see Advanced Usage. Under the hood, kokoro uses misaki, a G2P library at https://github.com/hexgrad/misaki Model SHA256 Hash: 496dba118d1a58f5f3db2efc88dbdc216e0483fc89fe6e47ee1f2c53f18ad1e4 Data: Kokoro was trained exclusively on permissive/non-copyrighted audio data and IPA…

Open weights apache-2.0

View model

Model · Text to speech

audio.cpp-gguf

Audio.cpp

This directory contains audio.cpp-native GGUF conversions of multiple speech models. These files are intended for use with audio.cpp. If you enjoy the project, please star audio.cpp on GitHub and this Hugging Face repository. For conversion details, supported layouts, direct-file loading, sidecar embedding, and the latest compatibility notes, see the audio.cpp GGUF guide: - https://github.com/0xShug0/audio.cpp/blob/main/docs/gguf.md!!! Converted and quantized packages are checked with automated metrics, but perceived quality can still differ for human listeners. Please validate the exact package, backend, and route to confirm the output is acceptable for your use case. The table lists the…

Open weights other audio.cpp

View model

Model · Text to speech

XTTS-v2

Coqui.ai

ⓍTTS is a Voice generation model that lets you clone voices into different languages by using just a quick 6-second audio clip. There is no need for an excessive amount of training data that spans countless hours. This is the same or similar model to what powers Coqui Studio and Coqui API. - Supports 17 languages. - Voice cloning with just a 6-second audio clip. - Emotion and style transfer by cloning. - Cross-language voice cloning. - Multi-lingual speech generation. - 24khz sampling rate. - 2 new languages; Hungarian and Korean - Architectural improvements for speaker conditioning. - Enables the use of multiple speaker references and interpolation between speakers. - Stability…

Open weights other coqui

View model

Model · Text to speech

chatterbox

Resemble AI

Chatterbox Multilingual V3 is the latest general-purpose multilingual TTS model in the Chatterbox family. It keeps the same 0.5B model size while improving speaker similarity, reducing hallucinations, and producing more natural, conversational speech across languages. V3 is designed for broad language coverage like V2, but with stronger stability and more expressive generation. It is the recommended multilingual model for users who want one voice cloning model that works across many languages. Try it in the Chatterbox Multilingual TTS V3 Space. Alongside V3, we are releasing the Single Language Pack: dedicated finetunes for priority languages where tighter quality control, stronger…

Open weights mit chatterbox

View model

Model · Text to speech

F5-TTS

Yushen CHEN

Download F5-TTS or E2 TTS and place under ckpts/ Paper: F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching

Open weights cc-by-nc-4.0 f5-tts

View model

Model · Text to speech

Kokoro-82M-v1.0-ONNX

ONNX Community

Kokoro is a frontier TTS model for its size of 82 million parameters (text in/audio out). First, install the kokoro-js library from NPM using: You can then generate speech as follows: Optionally, save the audio to a file: The model is resilient to quantization, enabling efficient high-quality speech synthesis at a fraction of the original model size.

Open weights apache-2.0 transformers.js

View model