Alt Flux & Collaborative Music Videos
My friend Ioan runs three YouTube channels for AI art:
Pixaroma (the main bread-and-butter one) focuses on ComfyUI workflows.
AI2Play which looks at alternate tools including nano banana, Google Whisk, ChatGPT, Midjourney, etc.
AltFlux, which has epic cinematic Music videos. Within AltFlux, there are two types of videos: (i) Short, action-packed videos that have a fast temp, and (ii) Scenic, hour-long, ambient ones that are suitable for meditation and relaxation
The challenge with short action-filled YouTube Music Videos:
It is very, very time-consuming to make a short, three- or four-minute AI music video. AI-generated videos are usually either 5 or 10 seconds long. You have to run many iterations because maybe half or two-thirds of the AI clips are unusable. They have some errors in them.
Then a few tries to create the appropriate music. Editing in software like Premiere Pro or CapCut to arrange the video segments in a chronological order that makes sense, each clip cut to the beat of the music, and the whole video color graded etc.
It can take anywhere from 8 to 20 hours for a short music video! And even when the channel is monetized, with such a short video, you can only have one ad at most to earn some revenue.
If the channel is not yet monetized, as is the case with AltFlux, it isn't easy to accrue enough Viewer hours to meet the monetization requirement. 1000 members, subscribers, and 4000 watch hours.
One solution: ambient videos.
These hour-long help set the mood for meditative contemplation, or they can be used as background music. A perfect example is when I saw a video of a scene in the woods with a fire crackling in the hospital waiting room. These would be useful in a spa or maybe early in the morning when you want to clear your mind. One criticism is that some people, not in the mood for meditative videos, view these ambient videos as low-effort "AI slop".
Another solution: Collaborative videos.
The collaborative videos were an idea proposed by my friend Hammel. The idea is for members of the Discord community to each provide short video clips, which Ioan can combine into a longer, more action-packed music video.
This helps when implemented as a weekly challenge. This allows the members, first of all, to practice building reps in video generation and to improve their skills.
Like Malcolm Gladwell's 10,000 hours or Thomas Edison's 1,000 ways that didn't work to invent a perfect light bulb, regular practice is crucial for success.
It is also beneficial to Ioan in the long run because it incorporates ideas he may not have considered. It lets the community develop their Video skills and saves him the time and credits to generate that many videos on his own.
Medieval Challenge AI Collab Project – Edition 1
I feel the first edition was a huge success.
Edition 2 will focus on Ancient Egypt.
My work-in-progress contribution towards the second edition.
I had made an initial attempt at cover art for Altflux using my QR Code Monster workflow in ComfyUI. I made a lot of variations —maybe about 100. I showed six variations to Ioan.
He proposed a better version using Google Whisk, generated with a prompt aided by ChatGPT.
A hyper-realistic 3D logo shaped exactly like the stylized triangular symbol with curved extensions, formed from intertwining serpents that perfectly follow the logo’s lines and geometry. Each snake has emerald-green scales with gold accents and gemstone-like reflections, their bodies merging seamlessly to create the structure of the emblem. The central intersection features a snake head with open fangs and glowing red eyes, radiating power. The scene is set against a vast epic mountain landscape at sunset, lightning in the sky, ancient temples carved into cliffs, rays of divine light breaking through storm clouds, mist rising from waterfalls, cinematic atmosphere, ultra-detailed textures, photorealistic lighting, dramatic depth of field, majestic fantasy tone, ultra high resolution.
This is the first blog post I think I've entirely dictated on the fly with WhisprFlow, without any AI-assisted editing. Let me know in the comments if you found this interesting or helpful.