A tiny guesthouse in Tuscany could never afford a cinematic ad.
So it imagined one instead.
A retired superhero, a monster truck, and the easiest parking in Montepulciano.
2'22", made solo. Sound on.
Parts one and two of the Christmas episode of "Villa Paola", my AI comedy series about a retired superhero in Tuscany.
Every hero was a kid once. This one was never given a toy. Now a whole village thinks he's Santa Claus, and he hears it as orders.
Half toon (minimax H3 MAX, 'cause I love the speed), half photoreal (Seedance 2.5), a Tuscan Christmas crooner made on Suno, all dialogue dubbed in post. One-man pipeline.
The night itself comes in December. Sound on.
The world's largest brokers is entering into the realm of the @avax 🔺
Charles Schwab, with $13.33T in client assets, just added $AVAX trading to Schwab Crypto.
Now, $13 Trillion of assets can be tradeble onchain right on Avalanche.
For once, Wall Street likes seeing red. 🔺
The only financial chain where Institutions loved their money parked and tradeable onchain
yep: going 480p is enough on most of the cases.
Btw, I don't know if it's avaiable elsewhere, but in Venice AI you can generate in seedance @480p as "DRAFT" mode, and they give you 7 days to upscale through Seedance. I don't know in which part of the generation they do the trick, but IF it's done natively, it's very interesting. Personally, I stick top 480p and that's it.
Halloween special of "Villa Paola", my AI comedy series about a retired superhero in Tuscany. Full episode, one-man pipeline, ~48h from script to upload.
Stack: MiniMax H3 for the toon prologue (open weights, ran some prompts locally), Seedance 2.5 for the photoreal part (omni-reference + multicut), Suno for the creepy children's choir, ElevenLabs for extra voices.
Stick around for the post-credits scene. Feedback welcome, especially on the multicut pacing.
There's no rule about this, at least in my experience. Personally I try to choose the best way (that comes to my mind) of "directing" a sequence in order to send a message, since with AI we can try and afford the fail (kinda). I let very little to the dialogues because I try to use less audio references as possible, since I've very poor experience in sound.
In any case I use as many words as possible in the prompt (thanks to Claude) because any of them create a border in latent space that reduces the space for the machine to hallucinate. The pace I prefer to give in editing, mix and matching what I've generated.
@EmiliOlivi_ai Yep: insane.
Not a good thing, imo, but it's the only affodable way to start working on your own ip. Expecially now that's post-or-die.
AI didn't make me free, made me solo.
@screenweaver_ai I gave the length shot by shot, considering the possible duration of the prompted action in real life. Never forced the timing or the pace. Then I trimmed the results in AE if needed.
Thank you!
Actually there is a lot of postproduction in After Effects in prder to have things matching and be consistent.
I gave up editing in premiere or in cupcut, now I do in AE even if it feels crazy: in a 90" short movie there may be up to 35-40 cuts that it isn't a big number, while the things to fix are always a lot, so I found this weird workflow.
@Fanfulladev Yep, but generating 30" at once exposes the creator to huge risks, since anything not considered in the prompt may amplify along that duration.