All Voice Lab
Ultra-Realistic AI Voices & Cloning
324 followers
Ultra-Realistic AI Voices & Cloning
324 followers
All Voice Lab is the AI voice platform for ultra-realistic TTS & voice cloning. Powered by SOTA MaskGCT 2.0 model. Multilingual, expressive audio for creators & devs.
Flowtica Scribe
Hi everyone!
Sharing All Voice Lab, a new AI voice platform aiming for ultra-realistic results in generation, cloning, and more. It's powered by their SOTA MaskGCT 2.0 model.
They offer a suite of tools focused on high-quality, expressive audio:
🗣️ Text-to-Speech: Aims for human-like emotion, intonation, and clarity across 30+ languages.
🎭 Voice Cloning: Claims "seconds" to create highly realistic clones capturing timbre, style, and emotion.
🎙️ Voice Changer: Lets you refine your own recordings using their voice library and controls.
🎬 Video Translation: Includes a one-stop workflow for translating and dubbing videos automatically.
Have tried it myself, and the quality is genuinely impressive. I tested the multilingual capabilities with some classic tongue twisters, and found it handled them remarkably well – surprisingly better than E11Labs or GPT-4o in those specific tests.🤯
The platform emphasizes quality and multilingual capabilities, aiming to serve creators, developers, and businesses. API access is also available.
DiffSense
Maybe we need a TTS benchmark? It's difficult to know what TTS is good. 11labs sound too audiobook ish. OpenAI has one gd TTS and the other voices are mehh. I guess its difficult to measure TTS quality. But it should be possible to create a baseline og some sort. Some sentences that TTS services often struggle with maybe?
Flowtica Scribe
@sentry_co Yes, evaluating TTS quality is hard. Human perception remains key for things like naturalness and emotion, particularly across languages. Plus, creating better benchmarks requires lots of diverse audio data.
Let's try this. I'm already a fan of ElevenLabs, if you manage to reach the quality levels of that, it's a great feat as cloning voice and some other stuff is paid there and you're providing it for free! Although I can't upload a voice sample, I need to get some special permission it seems. Is there a reason for that?
Other than that, it seems pretty solid! I'll just need to record some samples
Finally a good competitor for 11Labs. Congrats on the launch! Works great, really fast on the main app. I will keep using it for my projects!
The pricing structure is really nice for the user as well.
Good luck!
ran a quick test and it's one of the best AI voice generators I've used! Will definitely be back to test more
Tyce
Congrats on the launch! I tried to generate an audio from a french poem (because why not) and the voice was great, although some inconsistencies on the diction and pronunciation. The voice (I selected french) was definitely very natural though! Where not too far away from natural assisted conversations!
@zaczuo, this is incredibly well-built: especially impressed by how All Voice Lab handled complex multilingual inputs with clarity and rhythm. I am curious: how does the MaskGCT 2.0 model handle prosodic alignment in low-resource languages or languages with tonal structures? Does it use dynamic segmentation or a predefined acoustic mapping for maintaining emotional consistency across translations? Would love to geek out more on your voice cloning stack. :)