Pendrokar (Yanis L)

reacted to hexgrad's post with 👀 12 days ago

Post

4440

To Meta AI Research: I would like to fold ylacombe/expresso into the training mix of an Apache TTS model series. Can you relax the Expresso dataset license to CC-BY or more permissive?

Barring that, can I have an individual exception to train on the materials and distribute trained Apache models, without direct redistribution of the original files? Thanks!

CC (Expresso paper authors whose handles I could find on HF) @wnhsu @adavirro @bowenshi @itaigat @TalRemez @JadeCopet @hassid @felixkreuk @adiyoss @edupoux

posted an update 20 days ago

Post

1618

🏆 TTS Arena: Added the Orpheus-TTS and CSM TTS model Spaces to the Arena Fork
🏆 Pendrokar/TTS-Spaces-Arena

Orpheus has been able to get into the Top 5 of available models.

Orpheus-TTS 🚀: MohamedRashad/Orpheus-TTS
Conversational Speech Model🌱: sesame/csm-1b

posted an update about 1 month ago

Post

1803

TTS Arena: Added the Spark-TTS model Space to the Arena Fork:
🏆 Pendrokar/TTS-Spaces-Arena

Spark-TTS ⚡: thunnai/SparkTTS

Rerouted Microsoft Edge TTS and XTTSv2 to have them back at the Arena. The Edge Space had Gradio API disabled, though a HF Space is not needed since it contacts a Microsoft server anyway. No clue how long this API will work. A ZeroGPU space is now used for XTTSv2.

posted an update 2 months ago

Post

2502

TTS: Added both Zonos model Spaces to the Arena Fork:
🏆 Pendrokar/TTS-Spaces-Arena

Zonos 🌍: Steveeeeeeen/Zonos

IMHO while Zonos still has fewer amount of capabilities than xVASynth, I'm still giving it the first row at the TTS capabilities table due to IPA and insta-clone/zero-shot:
1️⃣ Pendrokar/open_tts_tracker

replied to Keltezaa's post 2 months ago

Did you not notice that each ZeroGPU space has 12 or was it 24 core server-type CPU? That is more powerful than what you get with a CPU-Upgrade space. And you get 10 for $9!!! A bargain!

reacted to hexgrad's post with 👍 2 months ago

Post

5771

I wrote an article about G2P: https://hf.co/blog/hexgrad/g2p

G2P is an underrated piece of small TTS models, like offensive linemen who do a bunch of work and get no credit.

Instead of relying on explicit G2P, larger speech models implicitly learn this task by eating many thousands of hours of audio data. They often use a 500M+ parameter LLM at the front to predict latent audio tokens over a learned codebook, then decode these tokens into audio.

Kokoro instead relies on G2P preprocessing, is 82M parameters, and thus needs less audio to learn. Because of this, we can cherrypick high fidelity audio for training data, and deliver solid speech for those voices. In turn, this excellent audio quality & lack of background noise helps explain why Kokoro is very competitive in single-voice TTS Arenas.

reacted to Xenova's post with 🔥 2 months ago

Post

13157

We did it. Kokoro TTS (v1.0) can now run 100% locally in your browser w/ WebGPU acceleration. Real-time text-to-speech without a server. ⚡️

Generate 10 seconds of speech in ~1 second for $0.

What will you build? 🔥
webml-community/kokoro-webgpu

The most difficult part was getting the model running in the first place, but the next steps are simple:
✂️ Implement sentence splitting, allowing for streamed responses
🌍 Multilingual support (only phonemization left)

Who wants to help?

11 replies

·

replied to their post 3 months ago

Which model out of the 8 models listed on my post?

posted an update 3 months ago

Post

3398

TTS: Added Kokoro v1, Parler Large, LlaSa 3B & MARS 6 TTS models to the Arena.
Pendrokar/TTS-Spaces-Arena

Also had added MaskGCT, GPT-SoVITS & OuteTTS a month ago. OuteTTS devs did say that is too early for it to be added to TTS Arenas.

Mars 5 does have a space with open weights models, but inference is way too slow (2 minutes+).

3 replies

·

reacted to fdaudens's post with 🔥 3 months ago

Post

3577

🎯 Kokoro TTS just hit v1.0! 🚀

Small but mighty: 82M parameters, runs locally, speaks multiple languages. The best part? It's Apache 2.0 licensed!
This could unlock so many possibilities ✨

Check it out: hexgrad/Kokoro-82M

1 reply

·

reacted to hexgrad's post with 🔥 3 months ago

Post

8507

hexgrad/Kokoro-82M got an upgrade! ⬆️ More voices, more languages, pip install kokoro, and still 82M parameters.

GitHub: https://github.com/hexgrad/kokoro
PyPI: https://pypi.org/project/kokoro/
Space: hexgrad/Kokoro-TTS

11 replies

·

replied to their post 3 months ago

After 4000 votes F5 TTS fell near the bottom of the leaderboard, I extracted some sample from Emilia. Let us see if that changes anything.

reacted to dylanebert's post with 🤗 3 months ago

Post

2032

🟦 New Image-to-3D model from Stability AI

stabilityai/stable-point-aware-3d

here's how it looks, with TRELLIS for comparison

reacted to hexgrad's post with 🔥 4 months ago

Post

20404

📣 Looking for labeled, high-quality synthetic audio/TTS data 📣 Have you been or are you currently calling API endpoints from OpenAI, ElevenLabs, etc? Do you have labeled audio data sitting around gathering dust? Let's talk! Join https://discord.gg/QuGxSWBfQy or comment down below.

If your data exceeds quantity & quality thresholds and is approved into the next hexgrad/Kokoro-82M training mix, and you permissively DM me the data under an effective Apache license, then I will DM back the corresponding voicepacks for YOUR data if/when the next Apache-licensed Kokoro base model drops.

What does this mean? If you've been calling closed-source TTS or audio API endpoints to:
- Build voice agents
- Make long-form audio, like audiobooks or podcasts
- Handle customer support, etc
Then YOU can contribute to the training mix and get useful artifacts in return. ❤️

More details at hexgrad/Kokoro-82M#21

21 replies

·

reacted to hexgrad's post with 🔥 4 months ago

Post

4062

Merry Christmas! 🎄 Open sourced a small TTS model at hexgrad/Kokoro-82M

2 replies

·

replied to hexgrad's post 5 months ago

The original Arena's threshold is at 700 votes. But I am sure Kokoro will hold the position. The voice quality actually sounds close to ElevenLabs.

But StyleTTS usually is not very emotional. So it will fail where Edge TTS does. The phrases where the voice has to be sad or angry. For example Parler Expresso was overly jolly.

reacted to hexgrad's post with 🔥 5 months ago

Post

3045

self.brag(): Kokoro finally got 300 votes in Pendrokar/TTS-Spaces-Arena after @Pendrokar was kind enough to add it 3 weeks ago.
Discounting the small sample size of votes, I think it is safe to say that hexgrad/Kokoro-TTS is currently a top 3 model among the contenders in that Arena. This is notable because:
- At 82M params, Kokoro is one of the smaller models in the Arena
- MeloTTS has 52M params
- F5 TTS has 330M params
- XTTSv2 has 467M params

5 replies

·

reacted to hexgrad's post with 👍 5 months ago

Post

1446

@Respair just dropped Tsukasa: frontier TTS in Japanese Respair/Tsukasa_Speech
It's expressive, punches way above its weight class, and supports voice cloning. Go check it out! 🚀
(Unmute the audio sample below after hitting play)

replied to their post 5 months ago

True, a sample from the original dataset would probably be the best. My attempt to try to fetch one from Emilia dataset was unsuccessful as HF dataset viewer can only show the German samples. Emilia's homepage has a ASMR-y example prompt given.

replied to their post 5 months ago

True about the narration style sample, but that still did not stop XTTS in surpassing F5. Both use the same sample.

Yanis L

AI & ML interests

Recent Activity

Organizations

Pendrokar's activity