Rayan @AskRayan

Cofounder CEO @PrunaAI pruna.ai Paris, France Joined July 2021

Tweets

763
Followers

336
Following

1K
Likes

1K

Rayan @AskRayan

7 hours ago

@PrunaAI lets get this in @ComfyUI

0 0 2 243 0

View Details

Today, we open-source PrunaVAED, a faster drop-in replacement decoder for video generation with LTX-2.3! - 𝗛𝗼𝘄 𝗲𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝘁 𝗶𝘀 𝗶𝘁? ~1.7-2.1x faster decoder & ~50% lower peak VRAM - 𝗛𝗼𝘄 𝗴𝗼𝗼𝗱 𝗶𝘀 𝗶𝘁? Near-original visual quality & no changes to the latent encoding - 𝗛𝗼𝘄 𝗱𝗼𝗲𝘀 𝗶𝘁 𝘄𝗼𝗿𝗸? We pruned, trained & distilled video decoders to maximize efficiency of the LTX-2.3 series from @Lightricks Want to try the fast open-source integration? Check it here on Diffusers: huggingface.co/PrunaAI/PrunaV… Want to try the fastest video generation endpoint? Check p-video here on API: docs.api.pruna.ai/guides/models/…

16 38 317 13K 332

View Details

Bertrand Charpentier @Bertrand_Charp

4 days ago

There are tons of challenges and research on ML efficiency. We maintain a human selection of the best papers on the topic!

Pruna AI @PrunaAI

5 days ago

Discover curated efficiency AI/ML research at CVPR 2026 via the Awesome AI Efficiency repo - We selected 20+ research papers tackling SOTA problems in efficiency. - We cover pruning, quantization, caching, distillation, and more 📚 Curated in Awesome AI Efficiency:

2 1 15 871 5

0 1 5 343 1

View Details

Pruna AI @PrunaAI

6 days ago

Your coding agent can generate images, video, and audio on Pruna. We packaged that into portable Agent Skills. Install the full suite once, start a new chat, and describe what you want in plain language. Your agent picks the right skills for the job. What's inside: • Guides for prompting and API safety • Tools for P-Image, P-Video, P-Image-Edit, and more • Workflows for music videos, avatars, explainers, and reels Install: export PRUNA_API_KEY="your_key" npx skills add PrunaAI/pruna-skills@pruna -y ⭐️ Skills: github.com/PrunaAI/pruna-…

5 8 27 1K 6

View Details

Pruna AI @PrunaAI

2 weeks ago

We will be releasing a new model soon ....

16 9 55 2K 0

View Details

Pruna AI @PrunaAI

2 weeks ago

New Research presented at @icmlconf 2026 SCALE: How reliable are quantized models? To answer this question, our team considered 2-, 3-, 4-, and 8-bit models using 6 quantization methods and studied their uncertainty and robustness metrics under 15 natural input perturbations. What are the key findings? • While accuracy scales monotonically with total bits, reliability scaling is nonlinear with a peak at 4-bit precision. • The 4-bit precision offers the best reliability–efficiency tradeoff. • Quantized models can produce more reliable uncertainty estimates under natural perturbations than their full-precision counterparts, while staying close on task accuracy! 📄 Paper: openreview.net/pdf?id=jehhhDH… ⭐️ Code: github.com/PrunaAI/quanti… 📊 Poster: drive.google.com/file/d/1mwWKyo…

0 4 16 1K 2

View Details

Pruna AI @PrunaAI

3 weeks ago

Our AI playground is the way to get to know our models for free! - P-Image: $0.005 & 0.6s per run. - P-Image-Edit: $0.01 & 0.9s per run. - P-Image-Upscale: $0.005 & 0.9s per run up to 128MP. - P-Image-Try-On: $0.015 & 2s per run. - P-Video: $0.10 & 5.6s per run. - P-Video-Avatar: $0.025/s & 6s per run. - P-Video-Animate: $0.03 & 10s per run. - P-Video-Replace: $0.03 & 3.58s per run. Try our model here: …ground-production-861e.up.railway.app Find more information here: pruna.ai/all-models Sign up for our API: dashboard.pruna.ai/login

7 3 36 4K 13

View Details

Pruna AI @PrunaAI

3 weeks ago

We'll be at Raise this week! If you're around, let's connect. Reach out or comment here if you'd like to meet up.

1 1 19 590 0

View Details

Pruna AI @PrunaAI

3 weeks ago

$2,000 Prize Pool + Extra rewards. Everyone can participate! 🚨 'MAKE IT TALK, MAKE IT VIRAL' is LIVE. P-Video-Avatar. Up to 18x faster than similar avatar tools. Full-body control. Audio support in 20+ languages. Your avatar is waiting for you! Create freely. Tell your story. Show what this model can really do.

14 17 71 3K 45

View Details

Pruna AI @PrunaAI

4 weeks ago

OpenAI, Google, xAI, BFL and .... Pruna AI are leading the price-quality Pareto front for the text-to-image at @arena This means that there will never be a better performing model on their benchmarks at a specific price point. We are honoured to be on this list along with giants in the AI space. The cool thing is that P-models are commonly the fastest models out there but we'll share more on that later. Explore the full leaderboard, including a dissection per category here: arena.ai/leaderboard/te…

3 3 23 1K 5

View Details

Rayan @AskRayan

4 weeks ago

that prune though... and new ones coming up 👀

Arena.ai @arena

4 weeks ago

Gemini 3.1 Flash Lite Image (Nano Banana 2 Lite) has entered the Text-to-Image Arena. It ranks #5 overall. Google's latest Gemini-3.1-Flash-Lite-Image lands on the Pareto frontier: scoring 1251 at just $0.034/image. This is near flagship quality at a fraction of the price. The

11 22 311 62K 55

0 0 1 73 0

View Details

Rayan @AskRayan

a month ago

@OpenRouter pruna models when? 👁️👁️

0 0 2 21 0

View Details

Rayan @AskRayan

a month ago

@johnAGI168 P-Video by @PrunaAI 's going strong by solving a completely different need with video gen in 5s, try it out :)

0 0 1 163 0

View Details

Rayan @AskRayan

a month ago

@LukePiette Dude congrats you're killing it!

1 0 2 194 0

View Details

Rayan @AskRayan

a month ago

@remilouf Agreeeeed

0 0 0 209 0

View Details

Pruna AI @PrunaAI

a month ago

Yesterday, we shipped the fastest and cheapest try-on model. Runware shipped a new P-Image-Try-On guide, dress a full outfit from separate flat-lay garment photos. • Role-based API: one person ref + up to 11 garment refs, the model places each item on the right body region • Full outfits in one call: hat, top, bottoms, shoes, bag, no composites or multi-pass workflows • Pose reference: restage the same person and look in a new stance without a new photoshoot • Turbo mode: faster generation at the same price for catalog-scale previews Ideal for fashion e-commerce, styling tools, and catalog teams building virtual fitting rooms. @runware 👉 Guide: runware.ai/docs/models/pr… 📚 API: docs.api.pruna.ai/guides/models/… 🧩 P-Video, P-Image, P-Image-Edit & more: pruna.ai/all-models

1 1 6 450 1

View Details

Pruna AI @PrunaAI

a month ago

P-Image-Try-On is 70% cheaper until June 21st! Upload a photo of a person + garment refs, and instantly get a virtual try-on image. Make sure to try it at any of our launch partners 👇

Pruna AI @PrunaAI

a month ago

P-Image-Try-On delivers the fastest and cheapest virtual try-on on the market. • Affordable: $0.015 for the first garment, $0.008 for each additional one • Multi-garment: fit up to 11 garment references onto a person photo • Quality mode for <2s per garment • Turbo mode for

15 11 68 25K 14

0 3 5 418 1

View Details

Irtefa @_irtefa

a month ago

These @PrunaAI models are fast!

Cloudflare Developers @CloudflareDev

a month ago

The entire model suite from @PrunaAI is now on AI Gateway. Create, upscale, or edit images and generate videos with avatars or animate characters. All delivered at competitive costs thanks to Pruna optimization.

4 8 56 7K 16

1 1 5 422 0

View Details

Cloudflare Developers @CloudflareDev

a month ago

P-Image-Try-On from @PrunaAI available on @Cloudflare AI Gateway - 70% off until June 21st. Upload a photo of a person + garment refs, and instantly get a virtual try-on image developers.cloudflare.com/ai/models/prun…