well, I just need a tech savvy and / or noob (a bit old days irix/unix understanding spirit would definitely help) Flamer who is willing to interview me for a life session on a google meet or skype or teams call for 1 hour or 2 or even a whole afternoon. The Topic seems to not really trigger a broader old school Flame Artist audience yet. But I am absolutely sure this will change within the next 30 days when people see what’s possible if they can think it.
2 Likes
You should do a logik live, looks like alot of interest
3 Likes
I did build an audio pipeline for ComfyUI that automatically generates and translates complete voice over podcasts perfectly in ANY main language out of just a simple idea.
The resulting WAV files are stunning perfect VO’s in 24bit, 48khz. This, my friends is absolute next level! I developed the nodes by myself.
6 Likes
I think subtitles are so yesterday. Now the new XteVision tool set creates complete VO voice tracks (24bit/48khz) in nearly any main language. Using Ollama as LLM host and auto-switching models gives me chills what’s possible already. My next implementation will be automatic timecode alignment for VO snippets. Stay tuned
4 Likes
If you want be a Super Human
3 Likes
I’m not sure what is being shown here other than text-to-speech.
2 Likes
There is of course a far bigger picture behind. The problem of our time is that we don’t have it. I build solutions that fully automate content generation (from any possible source: Web, News, Radio or your own idea) . And without any mistake the user get perfect translated multi-language 24bit 48khz audio out of it. Together with taglines that generate prompts for the visuals… TTS is only a small part of the chain. More and more languages can be added. The combination of llm-agents with Image/Video/Audio generation can and will be used to feed Channels for Agencies, Marketing departments, and what not. Flame Artist’s who incorporate that kind of service will have a good chance to get new gigs.
And I totally forgot: The Clients Data will stay inhouse! Here is a picture that shows how a datasource from the web get automatically prepared to use for a finnish client (even I do not understand the language:
altho there is too much legal exposure for me to ever do an official translation for a client (I always insist clients provide the exact language of any copy, offer or legal title) I do take this as an interesting possibility to have publicly available data fed into flame and have it drive graphic or compositing processes. I’m reminded of a cool project a co-worker did this type of thing with Houdini several years ago:
2 Likes
Although that looks cool id try focus on things more specific to regular flame tasks here. How it all integrates, what worlflows can help with compositing, automating tasks etc
1 Like
I try to express what I see. Sure thing … On the Flame Task. What is it? Who defines it? This perspective changed already on the relation of age of the Agency supervisors say 3 or 4 years ago. Discreet Flame back then was THE absolute answer as a complex tool to Artists Creativity and this was many years the holy grail.
Now, I like it or not… We are not the only magicians in the room anymore. Kids from the Gaming generation Z do not follow any of our rules. But they created a platform called ComfyUI. The UI part is just any web browser connecting nodes now as we have in batch or in Nuke or in Fusion. I do have a realtime re-lighting workflow just right now in front of me, that makes me wonder what’s faster and better: action or the GenZ realtime web browser solution. I can’t ignore this facts anymore. That if WE don’t adapt and get things done… we re obsolete. I just try every day to let Flame and any Comping solution out there holds the upper hand. The real question is not how to integrate AI into Flame. But to integrate Flame power in clients new pipelines. I am pretty emotional on that topic.
3 Likes
have a look into this TD meetup https://www.youtube.com/watch?v=31xJWD3cpoc
It’s a good overview of the subject and its potential.
2 Likes
Thank you. I am per definition not a nuke guru but it look fantastic. I had the opportunity to talk to and with an likeminded software developer on the very last sunday in a life zoom call who did exactly this implementation of ComfyUI into Nuke.
From what I get is: we could introduce the whole workflow part user-editable with ALL parameters as well node switches into flame as well. I am just not sure this is what the flamer want or need. Of course this opens up a whole new world of creating! Cool stuff, thanks a lot. So my main question now is: how about a InFlamebrowser window, where parameter change on ANY node can be made? Just like in batch or action? ComfyUI without the UI is just similar to “nothing real shake” was in the old days on cli level imho.
Thanks to @philm and Thom for the Live Talk last week about the demo and all kind of deeper requirement insights !!!
Good news: the Flame-ComfyUI VFX pipeline is now absolute easy to install, and works with all kind of EXR sequences… performant!!!. I tested on Flame 2025.2.x as well on older 2023.2. Now I only need to adjust the included workflows. I developed a bunch of nodes for ComfyUI to handle EXR Frame IO as we know it and like it and Flame UNDERSTAND it. From now, you can also send alpha matte (paint) for inpainting directly from flame to ComfyUI… and the ice on the cake is new smart sequence handling while using multiple flame seats on 1x ComfyUI server node. No more cleanup and archiving… Now it feels like a part of flame itself.
11 Likes
looking forward to a logik live on this - looks really cool @Xteve
4 Likes
Digging deep into Multichannel/Multilayer EXR… I did finally found a reliable way to AUTOMATICALLY list any EXR layer from ANY EXR source sequence with corresponding connection output to get it to use in processing properly. Combine that with a roundtrip from Flame… opens endless new possibilities in Multichannel/Multilayer EXR processing. In fact, you can define anything as a layer in EXR. Get your Motion Vectors beside positioning, posing, depht, whatever to work for you!
1 Like
Anyone here tried this? Any reviews?
1 Like
The standard EXR(2) incl. alpha matte send & return works fine. I released it couple of days ago.
The Multichannel Multilayer roundtrip is not released yet. I will hopefully go on a test-session with Thom next week on exactly this.
1 Like
1 .The MultiLayer EXR Loader reads in the sequence from any CG supplier
2. EXR Layer List does what the name say: it lists ALL available layers and provides an corresponding output connector
3. EXR Layer Gatherer bundles all available layers from the Layer List node.
Between List and Gathering… you can play around and modify all the render passes.
4. MultiLayer EXR Writer finally writes FLAME-CONFORM MultiChannel/Multilayer EXR Sequences in 16 or 32bit.
5. The Multi EXR let you use ANY layer/channel inside Flame 
3 Likes
Great job dude! Have a look into this that seems pretty cool. DiffusionLight implementation for ComfyUI<!-- --> - ComfyUI Cloud
Also if you need people to test things i’m available.
2 Likes