OuteAI

[OuteAI @OuteAI]

@zverhu Yes, you can create speakers in other languages that are supported. Follow this guide to create the speaker profile: https://t.co/mLYaFc0ZEc

[OuteAI @OuteAI]

New model OuteTTS-1.0-0.6B

- Based on Qwen-3 0.6B
- Apache 2.0 (free for commercial and personal use)
- Supports 14+ languages
- Adds batched inference, enabling fast audio generation for long inputs (~0.1–0.02 RTF)

📦 Model Weights: https://t.co/gn9dqNDNQc

[OuteAI @OuteAI]

@JUQ_AI Yes, you can definitely add a new language by further pre-training the model or fine-tuning it. Check out the Unsloth guide: https://t.co/O5kAgswekl "Oute-TTS (1B)"

[OuteAI @OuteAI]

@Prince_Canuma Thanks for adding the model! 🙌

[👋 Jan @jandotai]

OuteTTS v1.0: A multilingual speech synthesis + voice cloning model built on Llama3.2-1B.

Supports one-shot cloning (~10s audio), native text input (no romanization), and automatic word alignment.

Compact, fast, and fully local - runs on llama.cpp or EXL2. @OuteAI

[Chubby♨️ @kimmonismus]

OuteTTS 1.0: Upgrades in Quality, Cloning, and 20 Languages

- Best Performance: Generate audio around 42 seconds in a single run (approximately 8,192 tokens). It is recomended not to near the limits of this windows when generating. Usually, the best results are up to 7,000 tokens.

- Context Reduction with Speaker Reference: If the speaker reference is 10 seconds long, the effective context is reduced to approximately 32 seconds.

Github-link in comments

[OuteAI @OuteAI]

New OuteTTS model is here Llama-OuteTTS-1.0-1B

Major upgrades in speech synthesis & voice cloning, now with smoother processing and native multilingual support for 20 languages!

🔗 Full details on what's new & weights:
https://t.co/akF1QwgG8w

[OuteAI @OuteAI]

OuteTTS can now run directly in KoboldCPP!

🔗 Check it out: https://t.co/SGe5n9jxwI

[Vaibhav (VB) Srivastav @reach_vb]

Introducing OuteTTS 0.3 1B & 500M 🔥

> Zero shot voice cloning
> Multilingual (en, jp, ko, zh, fr, de)
> Trained on 20,000 hours of audio
> Powered by OLMo-1B & Qwen 2.5 0.5B
> Speed & emotion control

> Powered by HF grants 🤗

Open science ftw!

[OuteAI @OuteAI]

@sorayesh_semo Real-time results are definitely possible! With the GGUF variant, I've achieved around ~0.35 RTF (Real-Time Factor) on an RTX 4090.

[OuteAI @OuteAI]

🚀 OuteTTS 0.3 is live!

🔹 1B & 500M models
🔹 Enhanced naturalness and coherence with punctuation support.
🔹 Now supports 6 languages: EN, JP, KO, ZH, FR, DE.

Compatible backends: Transformers, LLaMA.cpp, ExLlamaV2
Models on HF: https://t.co/e1F5DRonN6

[OuteAI @OuteAI]

🎉 Demo space is now also available on Hugging Face: https://t.co/EVsKBhACnt

[OuteAI @OuteAI]

@carlos_moit At the moment we don't offer API, but keep an eye out for future updates.

[OuteAI @OuteAI]

Try OuteTTS in llama.cpp 👀

[Georgi Gerganov @ggerganov]

Open the pod bay doors, HAL.

[Xenova @xenovacom]

Introducing TTS WebGPU: The first ever text-to-speech web app built with WebGPU acceleration! 🔥

High-quality and natural speech generation that runs 100% locally in your browser, powered by OuteTTS and Transformers.js.🤗 Try it out yourself!

Demo + source code below 👇

[Chubby♨️ @kimmonismus]

OuteTTS-0.2-500M: A new and improved lightweight text-to-speech model

[Vaibhav (VB) Srivastav @reach_vb]

Smol TTS keeps getting better! Introducing OuteTTS v0.2 - 500M parameters, multilingual with voice cloning! 🔥

> Multilingual - English, Chinese, Korean & Japanese
> Cross platform inference w/ llama.cpp
> Zero-shot voice cloning
> Trained on 5 Billion audio tokens
> Qwen 2.5 0.5B LLM backbone
> Trained via HF GPU grants

Model weights on the hub, you can even run this on a Raspberry Pi! Go run, inference now! 🐐

[OuteAI @OuteAI]

OuteTTS-0.2-500M: Our new and improved lightweight text-to-speech model!

✨Improvements:
- Improved prompt following & output coherence
- More natural, fluid speech synthesis
- Multilingual: EN, CN, JP, KR
- Enhanced voice cloning

🤗 Try it on HF https://t.co/tfWQBzUxEC

[OuteAI @OuteAI]

@harambe_musk Not at the moment, but there’s nothing too special about the training. It’s basic pre-training, similar to how you would train any other language model for next token prediction.

[OuteAI @OuteAI]

🎶 OuteTTS-0.1-350M: Teaching Language Models to Speak Using Audio Tokens and Forced Alignment!
Our new experimental text-to-speech model with only 350M parameters achieves high-quality speech using a pure language model-based approach.

https://t.co/CmgOtVVroS

[OuteAI @OuteAI]

@reach_vb Thank you for covering our model!