🎙️ After serving millions of users through our text-to-speech platform, one need kept coming up: fine-grained AI speech editing - the ability to modify existing speech. Today, we’re open-sourcing PlayDiffusion, a diffusion-based inpainting model built for that exact purpose. Fully open source and ready for use in your own projects. 🔗 GitHub: github.com/playht/PlayDiffus… 🤗 Demo: huggingface.co/spaces/playht… ⚡ Fal: fal.ai/models/fal-ai/playai/…

Another example of the multiple TTS parallel pipelines pattern. Here's a voice AI agent that speaks both English and Arabic, using a specific model/voice for each language. These are @PlayAIOfficial voices. The STT, TTS, and LLM inference is all running on @GroqInc. (The LLM is LLama 4 Maverick.)

It's been a while since Google has recovered, and we no longer see any issues with authentication. Thank you for your patience. We're releasing a new feature for Studio users, keep an eye out 👀

Our systems are currently down due to an outage from Google, our underlying authentication infrastructure. We will keep you all closely updated.

We’ve written more about how the model works, its architecture, and training setup in our blog. If you're building in this space or want to contribute, we’d love to connect. More here: blog.play.ai/blog/play-diffu…

Want to try now? Our Speech Editor uses the PlayDiffusion model under the hood. Just upload audio/video, edit the auto-generated transcript, and get updated speech. Same voice, new words.

PlayDiffusion can also act as a becomes a highly efficient TTS system. Instead of generating tokens one-by-one like AR models, it predicts all tokens at once and refines them in ~20 denoising steps. That makes it up to 50× more efficient in generation steps with no loss in quality. 🗣️ Try it as a full TTS model on Fal: fal.ai/models/fal-ai/playai/…

Editing speech with autoregressive models is hard. Want to change one word? You either: - regenerate the whole sentence (and lose rhythm), - splice in a new segment (and get mismatched word boundaries or artifacts), - or start mid-way (and affect delivery). PlayDiffusion uses a non-AR diffusion model to inpaint audio, cleanly filling in masked regions while preserving context.

try it for yourself with the PlayAI Speech Editor today app.play.ht/speech-editor?ut…

if the godfather was a twitch streamer in 2025...

The Speech Editor is now available for free - we look forward to seeing a lot of fun content! app.play.ht/speech-editor/?u…

Whether it’s audio or video, Speech Editor lets you: - Fix stumbles, filler words, or awkward phrasing - Clean up interviews or voiceovers - Or even make someone say something ridiculous (for fun, of course)

Meet PlayAI Speech Editor. Now you can edit speech like documents. No need to re-record. No compromise in quality. Just precision edits powered by AI.

"Sterling as read by Slavoj" 🪄✨ @AskPlayAI @PlayAIOfficial

Congratulations to our partners @Lightricks on their new SOTA video model! Huge for the open source AI community and for AI creatives!

We’re excited to launch our new model — LTXV-13B! 💫 LTXV-13B is our most advanced open-source video model yet, and a major leap forward in quality, speed, and accessibility. The future of generative video is here — and it’s fast!

In the world of boring content, your voice is your most creative asset. It's how you build a successful brand. Those who invest in developing their unique voice identity today will have the upper hand. This is what PlayAI unlocks: creative freedom.

We built PlayAI to solve one problem: making high quality content creation accessible to all. Our complete audio AI suite now includes: • Voice Changer - transform any voice with 10-second samples • Audio Cleaner - studio-quality noise removal in one click • PlayNote - turn text into multi-voice podcasts automatically • AI Voice Agents - create interactive voice experiences The old days are over. Our products help you spend your time doing what's important: growing and scaling your brand. We're curious to hear what you think. ↓

Podcasters, teams, field reporters: Meet the one-tap mute button for real-world noise.