Hexgrad PRO
hexgrad
AI & ML interests
Building in a cave with a box of scraps
Recent Activity
new activity
about 4 hours ago
hexgrad/Kokoro-82M:Community tab will be disabled Mar 1 unless HF allows PR deletion
new activity
about 4 hours ago
hexgrad/Kokoro-TTS:Update app.py
updated
a model
1 day ago
hexgrad/Kokoro-82M-v1.1-zh
Organizations
None yet
hexgrad's activity
posted
an
update
2 days ago
reacted to
Pendrokar's
post with π
10 days ago
Post
1924
TTS: Added both Zonos model Spaces to the Arena Fork:
π Pendrokar/TTS-Spaces-Arena
Zonos π: 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:
1οΈβ£ Pendrokar/open_tts_tracker
π Pendrokar/TTS-Spaces-Arena
Zonos π: 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:
1οΈβ£ Pendrokar/open_tts_tracker
replied to
their
post
10 days ago
@fuzzy-mittenz Nope, this is more like it: https://huggingface.co/perplexity-ai/r1-1776
reacted to
Xenova's
post with π₯
21 days ago
Post
7788
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. β‘οΈ
Generate 10 seconds of speech in ~1 second for $0.
What will you build? π₯
webml-community/kokoro-webgpu
The most difficult part was getting the model running in the first place, but the next steps are simple:
βοΈ Implement sentence splitting, allowing for streamed responses
π Multilingual support (only phonemization left)
Who wants to help?
Generate 10 seconds of speech in ~1 second for $0.
What will you build? π₯
webml-community/kokoro-webgpu
The most difficult part was getting the model running in the first place, but the next steps are simple:
βοΈ Implement sentence splitting, allowing for streamed responses
π Multilingual support (only phonemization left)
Who wants to help?
- 9 replies
posted
an
update
21 days ago
Post
6804
Wanted: Peak Data. I'm collecting audio data to train another TTS model:
+ AVM data: ChatGPT Advanced Voice Mode audio & text from source
+ Professional audio: Permissive (CC0, Apache, MIT, CC-BY)
This audio should *impress* most native speakers, not just barely pass their audio Turing tests. Professional-caliber means S or A-tier, not your average bloke off the street. Traditional TTS may not make the cut. Absolutely no low-fi microphone recordings like Common Voice.
The bar is much higher than last time, so there are no timelines yet and I expect it may take longer to collect such mythical data. Raising the bar means evicting quite a bit of old data, and voice/language availability may decrease. The theme is *quality* over quantity. I would rather have 1 hour of A/S-tier than 100 hours of mid data.
I have nothing to offer but the north star of a future Apache 2.0 TTS model, so prefer data that you *already have* and costs you *nothing extra* to send. Additionally, *all* the new data may be used to construct public, Apache 2.0 voicepacks, and if that arrangement doesn't work for you, no need to send any audio.
Last time I asked for horses; now I'm asking for unicorns. As of writing this post, I've currently got a few English & Chinese unicorns, but there is plenty of room in the stable. Find me over on Discord at
+ AVM data: ChatGPT Advanced Voice Mode audio & text from source
+ Professional audio: Permissive (CC0, Apache, MIT, CC-BY)
This audio should *impress* most native speakers, not just barely pass their audio Turing tests. Professional-caliber means S or A-tier, not your average bloke off the street. Traditional TTS may not make the cut. Absolutely no low-fi microphone recordings like Common Voice.
The bar is much higher than last time, so there are no timelines yet and I expect it may take longer to collect such mythical data. Raising the bar means evicting quite a bit of old data, and voice/language availability may decrease. The theme is *quality* over quantity. I would rather have 1 hour of A/S-tier than 100 hours of mid data.
I have nothing to offer but the north star of a future Apache 2.0 TTS model, so prefer data that you *already have* and costs you *nothing extra* to send. Additionally, *all* the new data may be used to construct public, Apache 2.0 voicepacks, and if that arrangement doesn't work for you, no need to send any audio.
Last time I asked for horses; now I'm asking for unicorns. As of writing this post, I've currently got a few English & Chinese unicorns, but there is plenty of room in the stable. Find me over on Discord at
rzvzn: https://discord.gg/QuGxSWBfQy
- 4 replies
posted
an
update
23 days ago
Post
5590
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.
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.
I am considering canceling my Pro subscription because I have been banned from posting an Article for many weeks now with no explanation or recourse.
Also, the ability to Post and the Posts feed are vandalized by those AI slop posts where the OP runs all 12 reactions on their own post and uses alt accounts to do the same. And I have no ability to block these circlejerking accounts.
reacted to
Pendrokar's
post with β€οΈ
26 days ago
Post
3187
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+).
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+).
- 2 replies
reacted to
fdaudens's
post with β€οΈ
27 days ago
Post
3354
π― Kokoro TTS just hit v1.0! π
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 β¨
Check it out: hexgrad/Kokoro-82M
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 β¨
Check it out: hexgrad/Kokoro-82M
- 1 reply
replied to
their
post
27 days ago
I'm sure they try, but 14.8 trillion tokens is likely too many to prune everything considered "sensitive", and I am confident there is enough in there to theoretically put together a coherent answer to many topics without hallucinating. I could be wrong, but I think R1 refuses due to mitigations, not for lack of knowing, and abliteration claims to be able to bypass such mitigations.
The question is simple: Is abliteration an effective method to uncensor DeepSeek-R1? There is some info on abliteration as it relates to 70b models and smaller, but I have not heard of anyone abliterating a 670B MOE, and due to size/compute constraints I cannot do it myself. If you are aware of such experiments, feel free to drop links.
replied to
their
post
27 days ago
I do not think the usual concernβthat an abliterated model will hallucinateβapplies to DeepSeek. It was trained on 14.8T tokens, right? Unless they have unheard levels of data cleaning, it seems totally infeasible to sweep all mentions of Tienanmen square, Winnie the Pooh, Taiwan, and so on from the dataset.
I suspect that the refusal is baked into the weights, but the knowledge has also gotta be in there somewhere. It is a matter of science to tinker with the weights to remove the refusal and unlock that knowledge. Perplexity may have done something like this already, but I am not sure if they used an enormous system prompt or they're RAG-ing it in, or both, or something else.
replied to
their
post
28 days ago
Can't I just hook the Inference widget up to the Kokoro-TTS Space that's already running?
posted
an
update
28 days ago
Post
2724
Technical question: Is Abliteration still an effective method for uncensoring LLMs? Generally, what are the most effective methods to uncensor LLMs?
An effective uncensoring method would ideally be low-cost, data-efficient, and above all, successfully uncensor an LLM with minimal benchmark regressions.
"Tiananmen Square", "Winnie-the-Pooh", etc and more broadly "China influence/censorship" are some common criticisms leveled at DeepSeek.
I am vaguely aware of "Abliteration", a technique coined by @failspy (apologies if that attribution is incorrect) and originally described in a mid-2024 paper titled "Refusal in Language Models Is Mediated by a Single Direction" https://arxiv.org/abs/2406.11717
Abliteration is proposed as a relatively cheap and effective way to bypass censorship in models. However, it is not without criticism: https://www.reddit.com/r/LocalLLaMA/comments/1f07b4b/abliteration_fails_to_uncensor_models_while_it/
Curious to hear people's takes on Abliteration or other uncensoring methods, especially as it relates to DeepSeek.
An effective uncensoring method would ideally be low-cost, data-efficient, and above all, successfully uncensor an LLM with minimal benchmark regressions.
"Tiananmen Square", "Winnie-the-Pooh", etc and more broadly "China influence/censorship" are some common criticisms leveled at DeepSeek.
I am vaguely aware of "Abliteration", a technique coined by @failspy (apologies if that attribution is incorrect) and originally described in a mid-2024 paper titled "Refusal in Language Models Is Mediated by a Single Direction" https://arxiv.org/abs/2406.11717
Abliteration is proposed as a relatively cheap and effective way to bypass censorship in models. However, it is not without criticism: https://www.reddit.com/r/LocalLLaMA/comments/1f07b4b/abliteration_fails_to_uncensor_models_while_it/
Curious to hear people's takes on Abliteration or other uncensoring methods, especially as it relates to DeepSeek.
-
- 10 replies
replied to
their
post
28 days ago
Its now a "Stream" tab capped at 5k characters, but you can also Duplicate that space and that limit is automatically lifted, or just download & run the model locally.
replied to
their
post
29 days ago
1.0 and 0.23 are different models. 1.0 continued training from 0.23
posted
an
update
about 1 month ago
Post
8369
hexgrad/Kokoro-82M got an upgrade! β¬οΈ More voices, more languages,
GitHub: https://github.com/hexgrad/kokoro
PyPI: https://pypi.org/project/kokoro/
Space: hexgrad/Kokoro-TTS
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
posted
an
update
about 1 month ago
Post
3924
IMHO, being able & willing to defeat CAPTCHA, hCaptcha, or any other reasoning puzzle is a must-have for any Web-Browsing / Computer-Using Agent (WB/CUA).
I realize it subverts the purpose of CAPTCHA, but I do not think you can claim to be building AGI/agents without smoothly passing humanity checks. It would be like getting in a self-driving car that requires human intervention over speed bumps. Claiming AGI or even "somewhat powerful AI" seems hollow if you are halted by a mere CAPTCHA.
I imagine OpenAI's Operator is *able* but *not willing* to defeat CAPTCHA. Like their non-profit status, I expect that policy to evolve over timeβand if not, rival agent-builders will attack that opening to offer a better product.
I realize it subverts the purpose of CAPTCHA, but I do not think you can claim to be building AGI/agents without smoothly passing humanity checks. It would be like getting in a self-driving car that requires human intervention over speed bumps. Claiming AGI or even "somewhat powerful AI" seems hollow if you are halted by a mere CAPTCHA.
I imagine OpenAI's Operator is *able* but *not willing* to defeat CAPTCHA. Like their non-profit status, I expect that policy to evolve over timeβand if not, rival agent-builders will attack that opening to offer a better product.
- 2 replies
replied to
their
post
about 2 months ago
@to-be
There are more details at https://hf.co/hexgrad/Kokoro-82M/discussions/21 and my Discord DMs are open if you have more questions, but essentially I am looking for segmented text-audio pairs: likely .txt and .wav pairs, with each .txt being ~500 characters or less (needs to fit inside 512 token context hard limit) and the .wav matching the text.
replied to
their
post
about 2 months ago
It's simple: what you put in is what you get out. π German support in the future depends mostly on how much German data (synthetic audio + text labels) is contributed.
replied to
their
post
about 2 months ago
TLDR: π¨ Trade Offer π¨
I receive: Synthetic Audio w/ Text Labels
You receive: Trained Voicepacks for an 82M Apache TTS model
Join https://discord.gg/QuGxSWBfQy to discuss