Pruna AI (@PrunaAI) header
Pruna AI (@PrunaAI) profile picture

High-performance AI models that combine speed, quality, and specialization for key use cases. ๐ŸŒฑ Make AI accessible and sustainable for everyone!

Munich & Paris
Joined April 2023
Photo by Pruna AI (@PrunaAI)Photo by Pruna AI (@PrunaAI)Photo by Pruna AI (@PrunaAI)Photo by Pruna AI (@PrunaAI)Photo by Pruna AI (@PrunaAI)Photo by Pruna AI (@PrunaAI)Photo by Pruna AI (@PrunaAI)Photo by Pruna AI (@PrunaAI)Photo by Pruna AI (@PrunaAI)Photo by Pruna AI (@PrunaAI)
P-Video-Upscale just got faster and cheaper! โšก Introducing Turbo mode: faster video upscaling at 65% lower cost compared to Standard (up to 30 fps), with a slight trade-off in quality. Here's what you need to know: - ๐—ฃ๐—ฟ๐—ถ๐—ฐ๐—ถ๐—ป๐—ด: Starting at just $0.0035/s - ๐—ฆ๐—ฝ๐—ฒ๐—ฒ๐—ฑ: Faster processing across resolutions - ๐—ฅ๐—ฒ๐˜€๐—ผ๐—น๐˜‚๐˜๐—ถ๐—ผ๐—ป: 720p, 1080p, 2K, or 4K - ๐—™๐—ฟ๐—ฎ๐—บ๐—ฒ ๐—ฟ๐—ฎ๐˜๐—ฒ: Up to 60 fps, at no extra cost in Turbo mode Available via our inference partners @scenario_gg, @replicate, @WiroAI, @TellersAI, @eachlabs, @wavespeed_ai, Sociaro, and @runware. ๐Ÿ“š Model docs: docs.api.pruna.ai/guides/modโ€ฆ ๐Ÿงฉ Try it through the API or Playground: dashboard.pruna.ai/login
6
21
5
49
3,096
We put P-Video-Upscale Turbo Mode to the test! In our own benchmarks against Topaz Starlight, Topaz Proteus, and FLUX.3 Upscaler, P-Video-Upscale Turbo Mode is up to 18ร— faster and 14ร— cheaper at 1080p. Faster results, lower costs. Here's how they compare ๐Ÿ‘‡
P-Video-Upscale just got faster and cheaper! โšก Introducing Turbo mode: faster video upscaling at 65% lower cost compared to Standard (up to 30 fps), with a slight trade-off in quality. Here's what you need to know: - ๐—ฃ๐—ฟ๐—ถ๐—ฐ๐—ถ๐—ป๐—ด: Starting at just $0.0035/s - ๐—ฆ๐—ฝ๐—ฒ๐—ฒ๐—ฑ: Faster processing across resolutions - ๐—ฅ๐—ฒ๐˜€๐—ผ๐—น๐˜‚๐˜๐—ถ๐—ผ๐—ป: 720p, 1080p, 2K, or 4K - ๐—™๐—ฟ๐—ฎ๐—บ๐—ฒ ๐—ฟ๐—ฎ๐˜๐—ฒ: Up to 60 fps, at no extra cost in Turbo mode Available via our inference partners @scenario_gg, @replicate, @WiroAI, @TellersAI, @eachlabs, @wavespeed_ai, Sociaro, and @runware. ๐Ÿ“š Model docs: docs.api.pruna.ai/guides/modโ€ฆ ๐Ÿงฉ Try it through the API or Playground: dashboard.pruna.ai/login
3
14
538
Pruna AI retweeted
p-video-upscale now starts at $0.0035/s (yes 2 zeros) and is 2x faster than real time! It's also 6-18x faster than the competition and up to 14x cheaper!
P-Video-Upscale just got faster and cheaper! โšก Introducing Turbo mode: faster video upscaling at 65% lower cost compared to Standard (up to 30 fps), with a slight trade-off in quality. Here's what you need to know: - ๐—ฃ๐—ฟ๐—ถ๐—ฐ๐—ถ๐—ป๐—ด: Starting at just $0.0035/s - ๐—ฆ๐—ฝ๐—ฒ๐—ฒ๐—ฑ: Faster processing across resolutions - ๐—ฅ๐—ฒ๐˜€๐—ผ๐—น๐˜‚๐˜๐—ถ๐—ผ๐—ป: 720p, 1080p, 2K, or 4K - ๐—™๐—ฟ๐—ฎ๐—บ๐—ฒ ๐—ฟ๐—ฎ๐˜๐—ฒ: Up to 60 fps, at no extra cost in Turbo mode Available via our inference partners @scenario_gg, @replicate, @WiroAI, @TellersAI, @eachlabs, @wavespeed_ai, Sociaro, and @runware. ๐Ÿ“š Model docs: docs.api.pruna.ai/guides/modโ€ฆ ๐Ÿงฉ Try it through the API or Playground: dashboard.pruna.ai/login
3
3
21
513
How does P-Video-2-Pro-Reference perform across quality, speed, and cost? The latest P-Bench results show it sitting on all three Pareto frontiers. Across 7 reference-to-video models, the Ref2V Bench v0 results give: โ€ข Cost: #1 ยท ~$0.046 per output video second (including reference fees) โ€ข Quality: #2 ยท 1033 Elo (Rapidata) โ€ข Speed: #2 ยท 2.28s per output video second P-Video-2-Pro-Reference delivers the lowest cost in the benchmark, while ranking second in both human preference and generation speed. It sits on the preference ร— speed, preference ร— price, and speed ร— price Pareto frontiers. ๐Ÿฅ‡ Full benchmark: huggingface.co/spaces/PrunaAโ€ฆ
3
8
36
1,882
We open LoRAs to precisely remove objects (and their reflections) in seconds. It works with our previous distillation of Qwen 2.1 from @Alibaba_Qwen. Try it here - huggingface.co/PrunaAI/P-Q-Iโ€ฆ - huggingface.co/PrunaAI/P-Q-Iโ€ฆ
Today, we open-source P-Q-Image-2.1-Remove-Object, 3 LoRA adapters for Qwen-Image-2.1 by @Alibaba_Qwen. Remove the object you point at, along with its shadow or reflection, and change nothing else! - ๐—›๐—ผ๐˜„ ๐—ฒ๐—ณ๐—ณ๐—ถ๐—ฐ๐—ถ๐—ฒ๐—ป๐˜ ๐—ถ๐˜€ ๐—ถ๐˜? Remove objects in 6 steps instead of 40, with no CFG. Use the 6-step adapter as the default. An 8-step adapter is also available, and a 40-step adapter runs only on the base model, without the few-step model's drift. - ๐—›๐—ผ๐˜„ ๐—ด๐—ผ๐—ผ๐—ฑ ๐—ถ๐˜€ ๐—ถ๐˜? Shadows and reflections just go away along with the object, even when they sit outside the outline. The rest of the photo stays put and lines up with the input, so you can use the result as is. - ๐—›๐—ผ๐˜„ ๐—ฑ๐—ผ๐—ฒ๐˜€ ๐—ถ๐˜ ๐˜„๐—ผ๐—ฟ๐—ธ? The adapters load directly on top of Qwen-Image-2.1, so the pipeline stays unchanged. The few-step variants stack on our P-Q-Image-2.1 distillation LoRA. Mark the object with a rough red outline, or just name it in the instruction. ๐Ÿ‘‰ Blog: pruna.ai/blog/p-q-image-2.1-โ€ฆ ๐Ÿค— Open Weights: huggingface.co/PrunaAI/P-Q-Iโ€ฆ โšก๏ธ ComfyUI: huggingface.co/PrunaAI/P-Q-Iโ€ฆ
1
3
14
404
Today, we open-source P-Q-Image-2.1-Remove-Object, 3 LoRA adapters for Qwen-Image-2.1 by @Alibaba_Qwen. Remove the object you point at, along with its shadow or reflection, and change nothing else! - ๐—›๐—ผ๐˜„ ๐—ฒ๐—ณ๐—ณ๐—ถ๐—ฐ๐—ถ๐—ฒ๐—ป๐˜ ๐—ถ๐˜€ ๐—ถ๐˜? Remove objects in 6 steps instead of 40, with no CFG. Use the 6-step adapter as the default. An 8-step adapter is also available, and a 40-step adapter runs only on the base model, without the few-step model's drift. - ๐—›๐—ผ๐˜„ ๐—ด๐—ผ๐—ผ๐—ฑ ๐—ถ๐˜€ ๐—ถ๐˜? Shadows and reflections just go away along with the object, even when they sit outside the outline. The rest of the photo stays put and lines up with the input, so you can use the result as is. - ๐—›๐—ผ๐˜„ ๐—ฑ๐—ผ๐—ฒ๐˜€ ๐—ถ๐˜ ๐˜„๐—ผ๐—ฟ๐—ธ? The adapters load directly on top of Qwen-Image-2.1, so the pipeline stays unchanged. The few-step variants stack on our P-Q-Image-2.1 distillation LoRA. Mark the object with a rough red outline, or just name it in the instruction. ๐Ÿ‘‰ Blog: pruna.ai/blog/p-q-image-2.1-โ€ฆ ๐Ÿค— Open Weights: huggingface.co/PrunaAI/P-Q-Iโ€ฆ โšก๏ธ ComfyUI: huggingface.co/PrunaAI/P-Q-Iโ€ฆ
10
27
10
183
8,217
The Pruna Build Program is open for teams shipping real AI products. Everyone can try in Stage 0. Then, start with $300 in free credits after signup, and unlock up to $ 3,900 total as your usage grows. No heavy application: add billing info, build, and unlock the next stage when you hit usage milestones. Built for SaaS, API, and teams who care about cost, speed, and efficiency. You can also get early access to our P-Models, visibility, and extra support. Read how it works: buff.ly/fPacNsq Apply in 2 minutes: buff.ly/cstG31l Applications are reviewed monthly (third week of each month). ๐ŸŒ Website: pruna.ai/ ๐Ÿ“š Documentation: buff.ly/MxawQpQ ๐Ÿงฉ API: buff.ly/1l6X0XB
3
11
41
1,970
Pruna v0.3.5 is out! ๐Ÿš€ Here are a few highlights: - New FP8 quantization algorithms: static FP8, and time-aware FP8 for diffusion pipelines. - New MSE metric in the evaluation toolbox - Tighter security: safer checkpoint loading and a 7-day dependency cooldown - New tutorials for FP8 quantization and sustainability benchmarking ๐Ÿ‘‰ Read the notes: buff.ly/GnDidPi Thank you all for your contribution!
5
8
44
1,885
New challenge to celebrate the P-Video-2-Pro Series! $2,000 Prize Pool! ๐ŸŽญ ๐๐ˆ๐‚๐Š ๐“๐‡๐„ ๐‡๐Ž๐Ž๐Š. ๐๐‘๐Ž๐•๐„ ๐“๐‡๐„ ๐Œ๐Ž๐ƒ๐„๐‹. ๐—ฃ-๐—ฉ๐—ถ๐—ฑ๐—ฒ๐—ผ-๐Ÿฎ-๐—ฃ๐—ฟ๐—ผ ๐˜€๐—ฒ๐—ฟ๐—ถ๐—ฒ๐˜€. Generate the scene, lock the character, deliver the line, or recast the actor. Pick the model that fits the job. Cut a vertical microdrama hook that proves quality, speed, or price.
16
22
1
88
3,803
How to enter: 1. Create a 20-120 second vertical microdrama scene with one clear hook. Show your preferred model from the P-Video-2-Pro series and why it wins on quality, speed, or price. 2. Post on X or Instagram 3. Tag us and add #madebypruna 4. Submit your post link on @GimiApp: app.gimi.co/campaigns/pick-tโ€ฆ Use other tools or P-models to improve the final content quality, but donโ€™t use them to create misleading or deceptive content. Join our Discord to discuss and share about it: discord.com/invite/JFQmtFKCjโ€ฆ
2
4
460
Available until rewards run out. Donโ€™t miss it! Start creating now using the playground in the dashboard (~2.5K free generations after creating the profile): dashboard.pruna.ai/login
1
327
The @sinceaihq hackathon is kicking off, and @PrunaAI is a tech partner! ๐Ÿš€ Good luck to everyone building this weekend! Build some cool shit ๐Ÿ”ฅ
4
8
160
P-Video-2-Pro-Replace is top-tier in overall human ratings, at a third of the cost and 1.6ร— faster. - ๐๐ฎ๐š๐ฅ๐ข๐ญ๐ฒ: within 3% of the top-rated model at 480p - ๐‚๐จ๐ฌ๐ญ: 3.2ร— cheaper, just $0.05/s vs $0.16/s at 480p - ๐’๐ฉ๐ž๐ž๐: 1.6ร— faster than the next-fastest model, 6ร— faster than SkyReels V4 - #1 for ๐”๐†๐‚ ads at 768p, #1 and #2 for ๐ฅ๐ข๐ฉ ๐ฌ๐ฒ๐ง๐œ Across both Quality ร— Price and Quality ร— Speed, P-Video-2-Pro-Replace stands out in the high-quality, low-cost, fast-generation region. From UGC ads to music videos, swap characters without compromising on speed or budget. Benchmarked by @Rapidata. ๐Ÿงฉ API & Playground: buff.ly/OHtx3eM ๐Ÿ“š Documentation: buff.ly/bHQHKV0 ๐Ÿ“Š P-Bench: buff.ly/Fk54Z2K
7
4
29
2,331
Pruna AI retweeted
Two jars. Two worlds. Same glass. ๐Ÿซ™ I wanted to build worlds small enough to hold, and to look at them the way you look at something through fogged glass: carefully, and a little bit guilty for watching. Jar 1 & Jar 2 is a night market: amber bulbs, satay smoke, tiny silhouettes who never notice you. Jar 3 & Jar 4 is a forest that breathes light: teal moss, violet mist, fireflies drifting like thoughts. One warm and crowded, the other cool and alone. I wanted the camera to feel like a held breath moving from one to the other. Both were made with P-Image for the frames and P-Video for the motion, by @PrunaAI . The part I enjoyed most was choosing a single frame with the right light before letting it move. There are more jars on the shelf. Which world should I open next? A room with its own weather, cloth that remembers, or a sea in a teapot? ๐Ÿ‘‡ #PrunaAI #AIart #miniatureworld #PImage #PVideo
๐Ÿค– Made with AI
9
5
1
37
3,825
Pruna AI retweeted
P-Video-2-Pro-Replace by @PrunaAI is now on each::labs. Send a video and a photo, and get the same scene back with a new person in it. โ†’ Motion, camera and sound stay as filmed, with lips following the original audio โ†’ Replace one or two people, or just one person in a group โ†’ Clips up to 2 minutes, in up to 4K Try it now โคต
20
3
39
939
Pruna AI retweeted
Okay, this is actually pretty cool! Took a nighttime forest exploration video and gave the main character a completely new look using @PrunaAI P-Video-2-Pro-Replace. Just one reference image and a character swap, without touching the rest of the scene.
๐Ÿค Paid partnership ยท ๐Ÿค– Made with AI
Introducing P-Video-2-Pro-Replace, based on @MiniMax_AI H3! ๐ŸŽฌ Replace the people. Keep the scene. Swap one or two people in an existing video using reference images, while preserving the original motion, camera, framing, background, and audio And for launch, it's 50% off for the first two weeks, until November 19 at 6:00 PM CET! Here's what you need to know: - ๐—ฃ๐—ฟ๐—ถ๐—ฐ๐—ถ๐—ป๐—ด: Starting from $0.05/s at 480p - ๐—ฆ๐—ฝ๐—ฒ๐—ฒ๐—ฑ: ~1.5โ€“2 minutes for a 10-second video - ๐—ฅ๐—ฒ๐—ณ๐—ฒ๐—ฟ๐—ฒ๐—ป๐—ฐ๐—ฒ๐˜€: Replace up to two people using reference images - ๐——๐˜‚๐—ฟ๐—ฎ๐˜๐—ถ๐—ผ๐—ป: Supports videos up to 120 seconds Available via our inference partners @runware, @wavespeed_ai, Sociaro, @TellersAI, @WiroAI, @Segmind_ai, @replicate, and @eachlabs ๐Ÿ“š Model docs: docs.api.pruna.ai/guides/modโ€ฆ ๐Ÿงฉ Try the model through the API or Playground: dashboard.pruna.ai/login
4
5
2
24
3,214
Pruna AI retweeted
Swap people in a video while keeping the motion, camera, background, and audio untouched? ๐Ÿคฏ P-Video-2-Pro-Replace New model by @PrunaAI does just that. Up to 2 people, videos up to 120s, from $0.05/s, and it's 50% off until Nov 19. I've already test it ๐Ÿ‘‡
Introducing P-Video-2-Pro-Replace, based on @MiniMax_AI H3! ๐ŸŽฌ Replace the people. Keep the scene. Swap one or two people in an existing video using reference images, while preserving the original motion, camera, framing, background, and audio And for launch, it's 50% off for the first two weeks, until November 19 at 6:00 PM CET! Here's what you need to know: - ๐—ฃ๐—ฟ๐—ถ๐—ฐ๐—ถ๐—ป๐—ด: Starting from $0.05/s at 480p - ๐—ฆ๐—ฝ๐—ฒ๐—ฒ๐—ฑ: ~1.5โ€“2 minutes for a 10-second video - ๐—ฅ๐—ฒ๐—ณ๐—ฒ๐—ฟ๐—ฒ๐—ป๐—ฐ๐—ฒ๐˜€: Replace up to two people using reference images - ๐——๐˜‚๐—ฟ๐—ฎ๐˜๐—ถ๐—ผ๐—ป: Supports videos up to 120 seconds Available via our inference partners @runware, @wavespeed_ai, Sociaro, @TellersAI, @WiroAI, @Segmind_ai, @replicate, and @eachlabs ๐Ÿ“š Model docs: docs.api.pruna.ai/guides/modโ€ฆ ๐Ÿงฉ Try the model through the API or Playground: dashboard.pruna.ai/login
2
2
6
771
What if my monkey vlogger suddenly turned into a DINOSAUR? ๐Ÿ’€ I tested @PrunaAI P-Video-2-Pro-Replace, and the concept is wild One character swap, while keeping the original motion, camera movements, scene, and audio intact. No need to recreate the entire video from scratch. Just replace the character and let the madness continue! AI video editing is getting seriously interesting.
๐Ÿค Paid partnership ยท ๐Ÿค– Made with AI
Introducing P-Video-2-Pro-Replace, based on @MiniMax_AI H3! ๐ŸŽฌ Replace the people. Keep the scene. Swap one or two people in an existing video using reference images, while preserving the original motion, camera, framing, background, and audio And for launch, it's 50% off for the first two weeks, until November 19 at 6:00 PM CET! Here's what you need to know: - ๐—ฃ๐—ฟ๐—ถ๐—ฐ๐—ถ๐—ป๐—ด: Starting from $0.05/s at 480p - ๐—ฆ๐—ฝ๐—ฒ๐—ฒ๐—ฑ: ~1.5โ€“2 minutes for a 10-second video - ๐—ฅ๐—ฒ๐—ณ๐—ฒ๐—ฟ๐—ฒ๐—ป๐—ฐ๐—ฒ๐˜€: Replace up to two people using reference images - ๐——๐˜‚๐—ฟ๐—ฎ๐˜๐—ถ๐—ผ๐—ป: Supports videos up to 120 seconds Available via our inference partners @runware, @wavespeed_ai, Sociaro, @TellersAI, @WiroAI, @Segmind_ai, @replicate, and @eachlabs ๐Ÿ“š Model docs: docs.api.pruna.ai/guides/modโ€ฆ ๐Ÿงฉ Try the model through the API or Playground: dashboard.pruna.ai/login
23
5
35
1,484