Pepe Vilanova
@pepe_vilanova
Followers
2K
Following
4K
Media
459
Statuses
6K
Early believer in founders playing the long game. @JME_Ventures š @AprendoVC āš¼
š§ [email protected]
Joined August 2011
š Me morĆa de ganas por poder enseƱar esto š Hoy, por fin, te presento @AprendoVC šš§µ https://t.co/8N5Q0lpVn1
12
42
248
Ni al mƩdico.
š“ SĆ”nchez explica por quĆ© no felicitĆ³ a MarĆa Corina Machado: Ā«No me pronuncio sobre los Premios Nobel. No entro a valorarlo, respeto y mucho el trabajo que haceĀ». https://t.co/1kyhEMGnq2
214
4K
12K
Feels like Iām only touching 1% of whatās possible. If anyoneās up for sharing tips (or helping me not lose my mind), DM! š
0
0
0
Thereās no āecosystem glueā. To get one deliverable you jump between 4ā5 tools (ideation > generation > upscale > edit > animate > export > deliver).
1
0
0
Bad news š Latency kills creativity. When it takes minutes to see if a shot works, iteration speed plummets. Real-time previews would be awesome.
1
0
0
Good news š«” Character consistency is way better than I expected. Keeping the same face across shots is finally becoming reliable.
1
0
0
Wish š« (2) I was surprised that upscaling isnāt built into the workflow. It feels like such a core step that should just be seamless.
1
0
0
Wish š« In @higgsfield_ai (and others) you canāt properly give feedback on whether a result was good or bad. Feels like a huge missed opportunity for improving outputs over time. Instead, it feels like you have to reset the whole process every time.
1
0
0
Learning 2 Most suites drop you into a jungle of models. Figuring out which one actually works for your product is insanely costly. For images Iāve landed on @NanoBanana, for short videos on @Kling_ai, for longer shots on @Seedance_Pro. Just endless trial & error.
2
0
0
At the same time, weāre still really bad at prompting ā and thatās on us. It takes a lot of time (and frustration) to learn how to express what youāre actually imagining. Yes, even if you ask ChatGPT for helpš«µš¼
1
0
0
It definitely takes time before you start getting good, professional results. Thatās why so many people try it out, make a couple of thingsā¦ and then quit. Either you deep dive, or you end up a bit skeptical that it can actually be useful.
1
0
0
Learning 1. The output is not always as good as the prompt. Sometimes it nails it, sometimes the model justā¦ ignores you. Prompting feels less like ācontrolā and more like ānegotiationā.
1
0
0
The last few weeks Iāve been freaking out, trying (and suffering) with AI image & video generation for real use cases (which Iāll share little by little). Here are some first learnings and opinions that I have collected over the last few weeks š¤Ŗ
2
0
1
Excited to announce our ā¬2.4M round led by @JME_Ventures, with support from @bonsaipartners and existing investors. The funding will accelerate Raudaās voice AI agents, expand research and product teams and scale operations across Europe. More info š https://t.co/DNSYbEF3fK
1
4
10
After a short break, w're BACK ! š 10,000 MAUs and growing fast. Getting ready for a HUGE launch this September 2025. @altan_ai is going to become the best place to build autonomous software.
6
3
20
Animajās Next Frontier: How Sixte de Vauplane Plans to Reinvent Kidsā Media with AI via @founderspress
thefounderspress.com
When Sixte de Vauplane talks about storytelling, he doesnāt just talk about screens or scripts. He talks about pipelines, proprietary models, and the
0
0
1
