Final version of the most accurate AI generated DMT hallucinations so far. We're not there in terms of accuracy, detail, or coherency yet, but i bet we will be in 2-5 years! The 2hr 4k version is here: A guide on how I made this is in the replies.

1. First, I created the base images following my tutorial here: 2. Used these images within the new Luma model to make frame-to-frame morphs with "smoothly transforming shapeshifting morphing transformational intricate 3D shapes" as the prompt. Each new video used the final frame of the previous input as its first frame to create continuous morphing. 3. Increased the frame rate to 30fps in Topaz to make it compatible with OpenAI's Sora model. 4. Reduced filesize to under 100mb using Handbrake for Sora compatibility. Without these two steps, Sora won't accept your Luma inputs. Note: After they've been remixed, don't use your Topaz-interpolated Luma outputs in the final edit because they look jittery as hell. 5. Remixed the Luma outputs in Sora at custom level 4-5 with the same prompt. This is necessary because most Luma outputs get very incoherent during the morphing process. However, they have more creative, complex, and accurate psychedelic morphing - which is why we use it over Sora alone. 6. Used Topaz to upscale the Sora outputs to 4K to match the resolution of the Luma outputs. 7. In DaVinci Resolve, created two timelines: the bottom one for Luma outputs and the top one for Sora outputs. The Sora outputs are strategically faded in and out when the Luma outputs inevitably become very blurry and messy. 8. Looped the 4-minute sequence to create a 2-hour extended version. 9. ???? 10. PROFIT!!!

I am BEGGING you, please add the ability to include a previous video as context when using start/end image frame prompting for more fluid scene transitions. And also the ability to "remix" a segment of an output for improving quality when incoherency arises. 🥺 plz

Announcing: Our most advanced speech-to-text model goes beyond accuracy to capture the real-world complexity of human conversation and deliver reliable, source-of-truth audio data. Explore Universal-2 updates 👇

nothing like DMT. lacks the correct stereotypical geometry. you are still in Euclidian space.

Very nice! A phenomenal representation of hyperspace 🙇🏼‍♂️✨🌌🌈

quite accurate, bravo

Looks cool. Genuine question, how does this help or benefit someone?

it helps me cuz i think its neat and trippy and looking at it makes me go wooah

Woah yeah that’s close.

Best I've seen...but still a long way to go. Impossible to describe.

@alexgreycosm