I've been experimenting with Seedance 2.0 to create seamless extensions. I created this sequence called UBUMBE FOREST. Seamless extensions can actually be quite tricky to do and requires some precise magic with editing to connect them. There is no shortcut here, you must experiment until you find the sweet... spot to make the transition seamless. Ubumbe Forest - During a routine foraging trip deep in the Ubumbe forest in Tanzania, a father guides his daughter off the beaten path. Here, hidden from the rest of their tribe, she steps into a breathtaking new world and comes face-to-face with insects and creatures she has never seen before, even encountering a rare Peafox. 🟢Extending a scene however is fairly simple: - Generate your first sequence. - Upload that video as an Omni reference. - Type this at the start of the prompt "Naturally continue the movement and momentum from the previous clip. Keep the exact same art style, lighting, character appearance, and physics." and then explain what happens next in this sequence. - Congrats, you now have a scene extension that naturally flows from where the last video ended. However this is only the start of the puzzle, you need to join these clips so they're seamless. Seedance 2.0 is able to match the pacing, remember the characters and even their voices using this technique. The hard part is the extension but check my second post to see how I'm doing it in Capcut.show more

This is how I do a seamless extension, turn on the audio it's fully narrated. This requires some editing magic and experimentation until you find a sweet spot. If you have any other solid tips or even a definitive value for when you need to cut out a frame for a seamless extension, let me know. However this works well, you just need to experiment and find that sweet spot until it looks seamless. #DreaminaSeedance2 #DreaminaAI #DreaminaCPP #Seedance2

You should check Tom's tool.

Aha! Awesome share Kōda! Will def have a look at this! Much appreciated.

@aimikoda Thanks for sharing @aimikoda ! I hope it helps Travis - the tool is pretty simple - it detects and removes the overlapping frames, balances tone/contrast, and crossfades audio.

@MrDavids1 @aimikoda Try it out. I’ve open sourced the code so anyone can check it out, make suggestions or fork.

@osva1313 @aimikoda Much appreciated Tom! Will certainly have a look at this and thank you for sharing 🙏🏽

This is very helpful, Trav! Thank you so much! Are you African? Because I love how you love African content, and you constantly create them.

Thank you Stella and it's my pleasure! Yes, i'm born and raised in South Africa, still living in South Africa as well, i'm in Johannesburg. I love creating stories on my continent and I like referencing places I've visited as well and then I put my own fantasy spin on them. I personally don't see Africa being showcased too much in the world of AI. I love my continent :)

Yesss, I absolutely agree! Mahn, this is why I'm creating Afro-centric stories as well. Wow, I literally thought you lived in the US. I'm working on my first film too..

AMAZING! Very very keen to see what you're working on, will keep a close eye on your posts Stella. I've been a bit absent on here for a month but starting to get back into posting again. That first film you're going to create is going to feel like a major stepping stone and a project you'll remember forever! Just be careful, you might get very addicted and you won't want to stop creating films lol. It's amazing we can just tell all sorts of stories today with the help of AI. Have fun and all the best with your project! I see you're from Nigeria. I personally haven't visited but would love to one day! I have however tried some Nigerian food this side, that Jolloff rice and chicken is too good hahaha.

Aw, you're so kind! And yes, I agree, it's very addicting lol, and I genuinely believe that it'll be a stepping stone for greater things to come. Honestly, it is really amazing for real. I'm happy that we're early, and putting Africa on the map I've not visited SA too, and I've not had food from South Africa hehe. Yes, Jollof and Chicken is good, but if you want to blend in with the locals whenever you come, there are tons of local food you can eat! Whenever you get the chance or opportunity to visit, I'll be happy to be your guide hehe, show you the good spots and stuff! Come during “detty December", lol and you'll have the time of your life! Anyways, here's the trailer for my filmmm!! Let me know what you think about this, Trav

very interesting and useful! thank you

I heard they are sorting this out in 2.1 👀

That would be a dream come true if they can just give us a dedicated extend feature.

Thanks helpful Travis Thanks for sharing

Only my pleasure Amira! I am going to experiment some more with one take sequences. In this video I managed to extend 4 times as a continuous one take sequence. I wish they would just build a feature that handles seamless extensions.

yes four generations is the breaking point

Lovely concept and short! 🫶🫶and very smooth transitions too. I’m wondering if you experiencr quality loss or audio breaking for the extension? I tried several times for extension but the output always have breaking audio and overlapping audio

Thank you Jeff, yes, quality does degrade over a certain amount of generations, I think once you try and extend after 5 generations the quality degrades quite badly, especially if it's a continuous one take sequence. I strategically insert new shots/camera angles into the sequence to advance my narrative. It's unfortunately a limitation. I haven't had audio break to be honest, it usually extends perfectly matching the pacing and audio in the sequence. I've never experienced overlapping audio, that is really odd.

Ah I see. Got it. I’ll make some tweak and see. Once again, thanks for sharing awesome stuff. Lovely concept honestly, would love to see this in full 🫶🫶

Amazing! Also, have you connected Magnific to ChatGPT MCP yet? I know you use Claude already. It would be great if you can show us how to do it. Thank you!

Thank you SPASE. I have not tried this yet but it's on my cards 👍 Could be very useful!

I just connected it. They have a page on Magnific on how to do it: Now I gotta figure out how to use it :)

🔥🔥🔥

Thank you bro!

How did you get human face references in Seedance?

I have zero issues with faces. I use Seedance 2.0 omni on Dreamina, Runway, Hailuo and on Pixverse as well, no issues on my end.

in what program are you using as omni reference? Thanks.

I am using Omni on Dreamina, Runway, Hailuo and on Pixverse.

This is great! My only problem so far using Seedance 2.0 is the real face recognized error. It's insane! The faces were newly created, not famous. The work arounds take so long to prepare, trial and error, even storyboards have to be treated.

Thank you Fran! Yeah sometimes I think it can be hit or miss but my miss rate is very very low, I can work with all sorts of faces, even children as you can see from this example and this is across Dreamina, Runway, Pixverse and Hailuo, no problems at all on my end, I don't know if maybe I have less restrictions because of my region (South Africa)? It's probably the one aspect of using Seedance 2.0 that i've struggled with the least which is odd since a lot of people struggle with this. I must say, I do encounter face restrictions occasionally and it's mainly when i'm using a video as an omni reference but when it's just images as omni, I can practically generate anything (No celebrities of course).

I think its a regional issue yes. From what I've read and asked ChatGPT. I'm based in Ireland, so I don't know why this is a problem, maybe Europe has more restrictions or regulations? Actually the EU is famous for over regulations, what am I thinking? 😄

Been struggling with this, thanks Travis!

My pleasure! Takes some experimentation to get it right. I wish they just offered a seamless extension feature so we didn't have to do it manually.

Interesting how the editing ends up being harder than generating the scene itself

Its hard but once you get to that sweetspot you can repeat it forever

Indeed! Just requires some patience and careful observation, it's worth it in the end :)

This is dope.

Don’t you lose some quality using a video as reference?

You do, it's unfortunately something you have to deal with, I try and balance it by deciding when i've reached a point where the degradation is manageable and then I will insert other shots into my sequence strategically to continue the narrative.

Have you had generations with this prompt where it is completely not usable, or is it always on point? Not talking about the acting, but rather the "continue the movement and momentum from the previous clip" part. For example, do you get generations where it just starts at a completely random frame? Or the quality is completely different than the video reference?

Never, for me personally it will always continues exactly from where the video ended. There have been instances where sometimes it cuts to a different camera angle but 99% of the time it always continues from the end, I have never had issues with this and extending has been easy, it's joining them that is the challenging part. However, quality degradation will happen the longer you try and extend it.