Pinned post
THE END OF THE BEGINNING | THE RECKONING 002 LIVE SHOW
@Erikcason x @WalkerAmerica
x.com/i/broadcasts/2…
@Erikcason x @WalkerAmerica
x.com/i/broadcasts/2…
Walker⚡️
Verified account This account is verified. Learn more@WalkerAmerica
THE Bitcoin Podcast @TitcoinPodcast | @TheReckoning21 | better half: @carlabitcoin | clanker: @carlabotcoin
5.1K Following 83K Followers 38K Posts
Latest posts 5
MY AI WORKFLOW:
I ended up editing this video entirely in ChatGPT with Astra. First time doing so. Up until now, ChatGPT has been way worse than both Claude and Grok Bot in terms of music videos, but this time it surprised me… it took 8 versions but honestly version 1 was already quite solid, the next 7 were just fine tuning little details. Astra, Claude Opus 5.5 and Grok Bot had the exact same clip bank to pull from, and I did three versions each with Opus and Grok Bot, but they were just missing the mark by way too much (which, again, has been the opposite of my experience up until now).
It’s also worth noting that Claude is where I have all the skills/workflows built for @carlabotcoin, so Claude *should* always be superior, and has been up until now from a “one shot” perspective (except in a couple cases where Grok Bot absolutely crushed both Claude and ChatGPT, which was the case for the ZERODAY music video).
This is the first music video where @carlabitcoin did ZERO editing in Final Cut Pro and we let the clanker handle everything. Pleasantly surprised that Astra did such a good job. I’ve also obviously had each model create handoff files for the others so they’re all improving based on what I liked from each version the other models created and what I didn’t like from what they created.
For those curious, all initial stills *with* Carla were generated by Claude via @higgsfield, but from an Element not from her Soul ID. Claude + Higgsfield used Soul ID to create the “real life” reference image, which Astra turned into anime using its in-house image gen. Then Astra created 5 more character images which I fed to Claude/Higgsfield to create the Element to use for all the Carla shots (Seedance 2.5 for the lip sync lines and Kling 4.0 for everything else).
Astra did every single one of the non-Carla stills. We did it this way because both Claude and Grok struggled to get the anime style we wanted, and Astra nailed it on the first pass. Claude still handled the Kling 4.0 animations of the stills though.
Will continue to test ChatGPT vs Claude vs Grok Bot on each of our new Carla Botcoin music videos and keep y’all updated.
I ended up editing this video entirely in ChatGPT with Astra. First time doing so. Up until now, ChatGPT has been way worse than both Claude and Grok Bot in terms of music videos, but this time it surprised me… it took 8 versions but honestly version 1 was already quite solid, the next 7 were just fine tuning little details. Astra, Claude Opus 5.5 and Grok Bot had the exact same clip bank to pull from, and I did three versions each with Opus and Grok Bot, but they were just missing the mark by way too much (which, again, has been the opposite of my experience up until now).
It’s also worth noting that Claude is where I have all the skills/workflows built for @carlabotcoin, so Claude *should* always be superior, and has been up until now from a “one shot” perspective (except in a couple cases where Grok Bot absolutely crushed both Claude and ChatGPT, which was the case for the ZERODAY music video).
This is the first music video where @carlabitcoin did ZERO editing in Final Cut Pro and we let the clanker handle everything. Pleasantly surprised that Astra did such a good job. I’ve also obviously had each model create handoff files for the others so they’re all improving based on what I liked from each version the other models created and what I didn’t like from what they created.
For those curious, all initial stills *with* Carla were generated by Claude via @higgsfield, but from an Element not from her Soul ID. Claude + Higgsfield used Soul ID to create the “real life” reference image, which Astra turned into anime using its in-house image gen. Then Astra created 5 more character images which I fed to Claude/Higgsfield to create the Element to use for all the Carla shots (Seedance 2.5 for the lip sync lines and Kling 4.0 for everything else).
Astra did every single one of the non-Carla stills. We did it this way because both Claude and Grok struggled to get the anime style we wanted, and Astra nailed it on the first pass. Claude still handled the Kling 4.0 animations of the stills though.
Will continue to test ChatGPT vs Claude vs Grok Bot on each of our new Carla Botcoin music videos and keep y’all updated.
There is no red, there is no blue.
There is the state, and there is you.View quoted post on X
LIVE WITH @ERIKCASON AND OUR AI CLANKER CO-HOST, TALLY, ON @TheReckoning21
THE END OF THE BEGINNING
x.com/i/broadcasts/2…
THE END OF THE BEGINNING
x.com/i/broadcasts/2…
we are going to get millennial communism followed by zoomer fascism followed by alpha monarchy