Terry Orr

408 posts

Terry Orr banner
Terry Orr

Terry Orr

@TorrWeb

Electronic Engineer | Software Developer | Creative tinkerer | Tech, music, video & AI 🚀 YouTube: https://t.co/LcmPGWpDXr

Pacific Northwest Katılım Mayıs 2023
135 Takip Edilen140 Takipçiler
Dead Sun
Dead Sun@deadsun_0·
@DeepPromptCine With the For You feed being what it is now, we are less likely to be seen outside of the AI bubble. I'm seriously considering just posting videos on IG/TT and shutting down my YT since I'm in purgatory there. No reason to post videos on X other than sharing with AI community.
English
2
0
6
70
Dead Sun
Dead Sun@deadsun_0·
Started this account in October of last year. In the past 4 weeks alone, after 2 good performers, here's how my video views compare to my entire 9-month history of posting on X: Instagram: 5.02x (3.7 years worth of views on X) TikTok: 1.37x (1 year worth of views on X)
English
9
0
25
357
Max
Max@madgreek65·
For the second film in a row, I am making a long, detailed story without using visual storyboards. The results are great. I think the storyboards do more harm than good at times, especially for complex prompts. I prefer detailed shot instructions inside the prompt.
English
6
0
11
572
Terry Orr
Terry Orr@TorrWeb·
@SamJWasserman I keep working at it, but I have a long way to go. My output may look like slop, but the time and effort I put into it, begs otherwise. 😃
English
2
0
3
53
Sam Wasserman
Sam Wasserman@SamJWasserman·
A director's job is not to collect beautiful shots. It's to make choices add up: performance, blocking, lens, rhythm, subtext. AI video is crossing into the same territory. The winners won't be prompt decorators. They'll be directors.
English
4
3
51
1.5K
🟢DeepForest_Studio_AI🟢
🟢DeepForest_Studio_AI🟢@DeepForest_Inc·
Krea2-identity-edit LoRAがv1.2にアップデート🙌 ID保持・顔の類似度が改善され、キャラクターシート、顔交換、インペイント/アウトペイント、Try-on、高解像度パスにも対応🤩 専用のComfyUI-Krea2Editも同時更新。参照忠実度を調整する「ref_boost」も追加✨ リンクはリプ⬇️ #Krea2 #ComfyUI
🟢DeepForest_Studio_AI🟢@DeepForest_Inc

Krea2-identity-edit LoRAがバージョンアップ! v1.1 が来ていたので試してみました🙌 2枚のリファレンス画像から、2人を同じシーンに配置するテスト🤓 v1.0よりID保持は良くなった印象🙌 顔や服装の一貫性はどうでしょうか?改善の余地がありますが、素晴らしいLoRAですね🙂 #Krea2 #ComfyUI

日本語
3
8
57
9.8K
Terry Orr
Terry Orr@TorrWeb·
@ChrisGwinnLA I'm keeping my eye on it. I suggest post episode 2 when its ready, with episode 3 following in August.
English
1
0
2
33
Christopher Gwinn | Grindhouse Glitch
I am about 80% done with episode 2 of Wendy's Wonderful Dreamworld - just need to add a few scenes and fix a couple others. The first half, which is the same length as episode 1 (4 minutes) is in very good shape and could be ready in days. The second half might take a couple more weeks to finish. Question for my followers: would you rather that I post the first half as episode 2, then post the second half as episode 3 in August, or would you prefer to wait for a longer episode 2 in August?
Christopher Gwinn | Grindhouse Glitch@ChrisGwinnLA

Wendy's Wonderful Dreamworld: an AI Technicolor Fantasy (Part 1). A depressed young woman named Wendy, living a life of drudgery as a tutor in Victorian-era England, uses the power of her dreams to transport herself to a dazzling fantasy kingdom where romance and adventure awaits. Made with Seedance 2, Grok Imagine, Nano Banana, Elevenlabs, Suno, Topaz Astra, and Magnific. © 2026 Grindhouse Glitch. All Rights Reserved.

English
10
3
47
1.5K
Terry Orr
Terry Orr@TorrWeb·
@SamJWasserman I can't think of any new technology where that wasn't the case. Naysayers are a necessary ingredient of change.
English
0
0
2
11
Sam Wasserman
Sam Wasserman@SamJWasserman·
Articles like this will start popping up more and more until we stop paying attention or caring about it because it will just be expected as part of life, business and filmmaking. We no longer say oh this company used computers to make x, I promise I believe same will be for AI.
Variety@Variety

Roughly 300 programs across Netflix's library have used generative AI across their production process this year, the company revealed. “We are increasingly leveraging these tools to deliver higher quality output more quickly and at a lower cost than traditional methods,” the company said. “In some cases, productions would have had to leave out key shots and sequences in the absence of GenAI technology.” variety.com/2026/biz/news/…

English
5
7
26
1.4K
Calvetio Neralis
Calvetio Neralis@SanctuaireNord·
Good night from Skaervine Skaervine has a charm as raw as its mountains. Open hearths, crowded tables, laughter carried through the cold, and those peaks watching over every roof. A view beautiful enough to make us jealous of those who get to fall asleep beneath it.
English
3
0
6
196
Terry Orr
Terry Orr@TorrWeb·
@mix_buchi_ It looks good to me. LTX is sufficient to get the story across. Hopefully, it will update soon. In any case, probably other open-source models will appear before long. I'm committed to open source, so Seedance and other cloud models are not for me.
English
0
0
0
37
buchi_
buchi_@mix_buchi_·
便乗して生成してみました。 LTX-Director2を使用して、T2V、ComfyUI内で完結。 いまいちプロンプトの効きが不完全だが、それなりの映像が手軽に生成出来るのかな、と。 あと、ComfyUIが0.28.0、torchが2.13.0になって生成がかなり速くなった。この映像(1280x768、15s)で生成126s。 #LTX23 #AImovie
Photogenic Weekend@PhotogenicWeekE

まぁまぁうまく行った例(笑)

日本語
3
0
22
3.2K
Terry Orr
Terry Orr@TorrWeb·
@xiaofan1233 @PhotogenicWeekE Ignored perhaps by everyone who can afford Seedance. In the meantime, there are 1000's who can't afford the high subscription costs.
English
0
0
0
7
Xiaofan
Xiaofan@xiaofan1233·
@PhotogenicWeekE Even if you gave the LTX 2.3 team ten years, they still wouldn't be able to achieve the results seen in Seednace 2.0. The model data for LTX is abysmal; currently, LTX 2.3 relies entirely on various LoRAs to stay afloat—otherwise, it would certainly be completely ignored.
English
1
0
0
28
Terry Orr
Terry Orr@TorrWeb·
@PhotogenicWeekE A poor story with Seedance is worse than a good story with LTX. At least that's my position. As for going toe-to-toe, it sure as hell beats Seedance on pricing. For some of us, that makes the difference between telling our story or not.
English
0
0
0
25
Sam Wasserman
Sam Wasserman@SamJWasserman·
I hate being sick. Losing hours and days where can’t get work done drives me crazy. Worst part is stressing about it makes it take longer to get better. Frustrating cycle. Anyways forced to get some rest. Be back soon.
English
7
1
22
653
ZOOT
ZOOT@LittleTinRobot·
帰還
日本語
8
40
262
17.4K
Terry Orr
Terry Orr@TorrWeb·
It's still pretty rough in spots. I consider this a first cut. My goal is to finish episode one and then go back and redo selected clips before moving on. Parts 1-3 take up around 4.5 minutes total. I estimate that the full Episode 1 will require somewhere around 8-10 minutes.
English
0
0
0
28
Terry Orr
Terry Orr@TorrWeb·
THE RIPENING: Episode 1 - Part 3. He has no title, no origin, and no equal. Marcus built the Curators from the shadows and lets others wear the authority. In Los Alamos, he meets with the local administrator to discuss a stalled research project, and the daughter of the man leading it. Meanwhile, that daughter is just trying to get through an ordinary evening, until a phone call changes what "ordinary" means. Created using open-source LTX 2.3. @ltx_io @ComfyUI
English
1
0
2
72
Terry Orr
Terry Orr@TorrWeb·
Free open-source voice cloning alternative to ElevenLabs. Maintaining character voice consistency across video clips is tough. This ComfyUI workflow fixes it by converting source audio to a perfectly synced character voice. Local • Unlimited • High quality (ChatterBox TTS)
Terry Orr tweet media
English
1
1
2
97
ArtemLutai
ArtemLutai@David_from_mars·
@TorrWeb ConvRot can actually be faster than FP8 on modern NVIDIA GPUs because its kernels are highly optimized. It offers an excellent balance of speed, VRAM usage, and image quality, making it one of the best quantization formats currently available for ComfyUI.
English
1
0
2
85
Max
Max@madgreek65·
Screw mobile formats. I am aiming for TV and Computers/Tablets. 16:9 all day for films
Max tweet media
English
4
0
11
227
Terry Orr
Terry Orr@TorrWeb·
It really annoys me when the AI tells me that I should call it a night because I've been at it for too long. Especially since I've only been at it for a couple of hours! 🤣
English
2
0
5
130