Monique Pryce

My goal with AI is to make it ✨i n v i s i b l e✨ Every decision I make, from prompting to selecting which generation actually works, is in pursuit of this aim. AI will always have imperfections, but the one thing I refuse to compromise on is whether my characters are giving a believable performance or not. Because imo, that's the one thing that can make people forget they’re watching AI at all. Take the attached clip for example: Watch Karl right after Beethoven says “Well?” He tosses his pen back into the holder and gives a look that says “brother, can you PULEAASE stop talking!” That's something that I prompted for specifically. [He tosses his quill down, irritated, and turns his head to the side to look at his brother, a little puzzled at what has got into him, and answers him. He asks line 1 as a real question, dry, mildly exasperated, as if to say, "What's up with you?"] Those details were enough to yield a performance that, to me, really felt like an older brother pestering his little brother. And notice how I'm saying "to me." Because I can't always tell you why I'm choosing one take over another. The reasoning behind my choice could very well come down to something as small as “he didn’t exhale fast enough.” I'm just letting my intuition be my guide. I could hand you my exact prompts, character refs and environments, and you’d still end up with a different scene from mine due to a difference in our intuition. But that's kinda the point, isn't it? Anyone can copy prompts and JuSt pUSh A bUtTOn until thy kingdom come, but knowing which takes to keep is the actual craft behind all of this.

Time to give y'all an updated AI toolstack! Here's what I use on the regular as of Q4 2026: -ˋˏ IMAGES ˎˊ- NBP - Still my go-to for environment references / character outfit changes. Also, i've been using the 4k option for upscaling (I've largely stopped using the legacy Magnific/Topaz labs.) Seedream 5 - Hands down, the BEST model for accurate character likenesses if you're working from a reference image (painting, death/life mask, photograph, etc) Midjourney - Still the best for unique styles. I use it all the time in cinematic projects for environmental plates. Everything comes out of a custom profile that I trained. FLUX 2 MAX - This one is new for me. I use it for upscaling environment plates. It adds perfect textures for everything, which is especially useful when combined with GPT image to get rid of all the friggin' artifacts. GPT image 2.6 sunburst - Surgical image editing. I never use raw images because it just has the weird brown tone and wayyyy too many artifacts (so annoying). Recraft v4.1 - logo generation Claude driving Blender - for environment layouts (I'll explain this more in another post.) Claude design - for quick thumbnails. Photoshop - For compositing/editing -ˋˏ VIDEO ˎˊ- Seedance 2.5 - My main video generator Kling 3 - For environmental/establishing shots with no humans Gemini Omni Flash - For generating sound FX and adding elements to talking head videos. CLAUDE CODE (driving Tesseract, working in premiere pro and After effects) - This has radically changed my editing work these last two months. I use it for everything from rough cuts to final polish. It also drives Blender for more impressive looking b-roll. I love it so much. -ˋˏ AUDIO ˎˊ- Elevenlabs - for professional voice cloning in particular. Minimax - for dialogue correction. Minimax is leagues above elevenlabs for this purpose. 11Labs tends to clean up audio and alters clones too much (with the exception of PVC). Problem is, there's not always gonna be the opportunity for PVC. But the instant voice cloning in Minimax is a TRUE instant clone, retaining things like room acoustics, accents, and background noise, so that when you patch audio in your video, the blend will be SEAMLESS! Sync . so - for correcting lipsync after correcting audio. Adobe podcast - for quick and easy audio cleanup. -ˋˏ PLATFORMS ˎˊ- Sume - My main platform these days. More cost effective than the other platforms I've used, and you can see the actual cost per generation vs. credits. Aaaaaand, NO SUBSCRIPTION COST! Magnific - backup platform. it has less moderation, which is great. I hate getting randomly flagged for NSFW even though it's nothing of the sort. And I love using nodes. Higgsfield - I'm trying to move away from this platform, as i'm just tired of some of the underhanded tactics. My annual subscription expires this month. After that, i'll be fully with Sume (and maybe Magnific depending on what that black friday deal looks like...). But i have to admit that some of the features are bomb... ComfyUI - I started experimenting with it, but i'm having a lot of trouble. 😅 I'm sticking with it only because I know how powerful it is once you get the hang of it. And it's about time I learned how to run things locally. -ˋˏ EDITING/PRODUCTION ASSISTANCE ˎˊ- Claude - My copilot / assistant producer. Covers prompts, troubleshooting, and everything else. It's great that we can now send mp4s to claude to "watch" as it makes reviewing soooooo much easier! And like I mentioned, I have it edit a whole lot as well. Wispr - For dictation. I practically never type anything anymore. Ain't nobody got the time fo dat! Premiere Pro - Main editing software After Effects - For masking Descript - for screen recording and captions (though Claude has been able to handle a lot of the captions lately.) And I think that's just about it. Hopefully I didn't forget anything... If you want me to elaborate on any of these, lmk!

I'm 100% convinced that more details = better performances from your AI actors. I'm talking about giving detailed instructions on how they should breathe, blink, swallow, chew, step, smile, etc. AND explaining the underlying emotion behind it. Yes, it might seem a little neurotic, but I ALWAYS end up with the exact emotion I want this way. It might take a few tries to find the right wording, but the winning prompt is always a highly detailed one. Here's an example from a previous project. This shot takes place during a court proceeding. And I'm sure that you'd be able to tell who's the defendant vs. the plaintiff just based on the emotions alone. For the man's closeup, here's the part of the prompt that did the heavy lifting: He stares straight ahead, stone-faced. His jaw is set tight. He is trying his best to project calm but his body betrays him. He hyperfixates on something out in the distance, his eyes do not move and he does not blink. He moistens his lips, swallows once, then the tension in his jaw resets. His nostrils flare slightly with a controlled exhale. His entire demeanor is that of someone hearing something he disagrees with but cannot respond to. And for the woman: She is barely holding it together. Her nostrils flare slightly as she takes a slow, controlled breath in through her nose. Her throat moves as she swallows once. Her eyes glisten but she blinks once, slowly, and deliberately like someone forcing the tears back. After she blinks her gaze drops — her eyes lower, her chin dips, her whole head tilts downward until she is looking at her own lap. She takes another breath — this one shakier than the first, her chest rising and falling visibly. She keeps her head down, eyes cast toward her hands. She is a woman who cannot watch anymore. Make note of how I used anchor gestures (“like someone forcing the tears back,” “someone hearing something he disagrees with but cannot respond to”) and bodily actions (“slow, controlled breath,” “moistens his lips, swallows once,” “the tension in his jaw resets”). Also notice how these aren't big, dramatic emotions but more subtle ones where you'd feel as though you can read what's on the person's mind. And for me, that's what makes a performance feel truly convincing. Aaaaand I should also mention that this is an old project from way back in April, made with Kling 3, before I started using my current prompt structure, which includes even more details and context. So yes. If you think you're detailed enough but your outputs don't feel convincing... ....you know what to try next!

To answer your questions: - No, there wasn't any upscaling involved. - This was generated in Seedance 2.5 at 720p - This clip has four total generations edited together. - It took 5-7 tries each to get the clips that were ultimately selected for this edit. - I used a midjourney profile trained on historical cinematic stills to generate the environment reference for this scene. (NBP and other models gave me generic looking results.) - I generated the character images using NBP/ Seedream 5 and historical references. - Claude is my assistant producer. - Contrary to what Bytedance published in their Seedance prompt docs, I'm 100% convinced that highly detailed, elaborate prompts make a big difference. My prompts for these were like essays. - "splitting generations in two" just means to give your generations room to breath. So if there are 6 lines split between two characters, rather than putting that all into one 30s generation, make it one 30s + one 15s generation to give your characters space to react, process, think, and all the little things that you'd detect through body language. - Yes, I've shared the prompts for this one. They're in the @DecimaVault Skool community. (and there's also a full walkthrough of my generation process on there right now, for those who are curious.) - For video generations I use @sumelabs primarily because their seedance costs are relatively cheap, you can see the price per generation rather than credits, and the UI is intuitive. - No, I did not use starting images. This was all just character refs (@sumelabs can y'all PLEASE add an option to create REAL elements with multiple angles??) - Character consistency is easy. Just create images of your character from various angles (my set always has front, quarter, profile, and full body) and attach them as references in your prompt. Or create an element with them and reference them within the prompt. Trust me, it isn't complicated at all. - For maintaining the same character voice, I've found that as long as you use the same reference images for the character, the voice will remain the same. I realized that with Veo3, and it remains true today. - I do NOT care about the minor inconsistencies and imperfections. If 95% of people won't notice upon first glance, then it's a go. (You're not gonna get absolute perfection with AI, and you shouldn't expect to. If you want 100% accuracy, go the traditional route.) - I haven't used driving performances for AI gens, but I'll be experimenting with it in the future. I have a hunch that I'll need it for future scenes. This one was text to video (with reference images of course) - No, this movie isn't out, and it isn't even close to being finished. I have it slated for March 26th of next year (the 200th anniversary of Beethoven's death). So no, you cannot watch it. 😜 - My budget for the film is about $5k. I've spent about $210 on the 5 minutes i have on the timeline right now. $5k is the max of what I'd be able to manage on my own, with just bare-bones personal expenses for the next few months, and a couple extra AI ad jobs on the side. (tho I do have a plan for fundraising in a few months. Everyone in my classical music network is gonna get a call from me come January. LOLOLOL) - The audio is native, but I patched certain gibberish words/phrases and corrected the lip sync. And btw, I'm well aware that STEEEPHAN was pronounced wrong (@suddenlyFleck flagged it in his review), and I'm trying to figure out the best way to fix it without ruining the acoustics. So German speakers...y'all can stop hounding me. I get it. 🤣 - Yes, I'll ultimately have native German speakers review the final drafts, and make the necessary corrections. - None of this production process is automated. I have my hands in every single step. AI can execute, but only I can make the final judgement call (taste is everything, right?) - No, I don't have a film background. My background is in classical music. - I've been in the AI space since late 2022. - Yes, I do this full time in different capacities. My main gig is working for @NetworkChuck as an AI-assisted editor. - I'm more than happy to share what I know in the DMs, so don't be shy! - Beethoven is my hero, but Scarlatti is the composer I liszten to the most. ....and i think that's it! Honestly, I'm feeling a tad bit overwhelmed by the amount of DMs and emails I've been getting (as if I wasn't already TERRIBLE at responding in a timely manner), but since I keep getting the same questions over and over, I just figured I'd make a post. And if you have a different question about workflows or whatever, feel free to DM and I'll get back to you...eventually...

If your AI characters sound fake, the fix is almost always in how you write and DIRECT the dialogue in your prompt. I recently started on the dialogue-heavy scenes for my AI feature film about Beethoven, and too many of the lines kept coming out wrong. They’d sound too theatrical, too monotone, or too much like a practice read-through between actors. (all this despite the fact that Seedance 2.5 is fantastic when it comes to realism.) So through a whole lot of trial and error, I've found that these 5 things made the biggest impact on the quality of the dialogue delivery in my generations: 1. Paying close attention to the written punctuation used. 2. Breaking long lines into shorter ones, with intentional pauses. 3. Using bodily actions as punctuation (this one gave the biggest improvement BY FAR! In the attached clip, note the sniffles in between Beethoven's lines. That alone made the delivery feel much more natural.) 4. Using "anchor gestures" (he lifts his finger as if to say "one sec") 5. Just split your generations in two (stupid simple, but it really makes a difference.) So if your characters keep sounding like they're reading from a teleprompter, give this a try!

AI editing is actually a BLESSING!! Before I went to bed, I gave Claude a thorough description of a b-roll sequence I had in mind for this segment. This morning I woke up to that sequence fully rendered and placed on my premiere pro timeline. My idea, perfectly executed while I slept. And y'all are COMPLAINING about this tech??? I just don't get it....

My vision of creating a feature length film about my favorite composer is taking shape piece by piece! It's so surreal to think that what was a pipe dream a decade ago is now manifesting itself in reality. God bless AI. For real. #thelastdaysofbeethoven

My first agent-assisted long form edit is now a 1of10. For the last few weeks I've been knee-deep in trying to leverage AI agents for editing. I saw this challenge going around where you never touch the timeline. You just give your agents instructions and let them edit. I tried the challenge for this edit....and I didn't last very long at all. lol Instead, the process looked like this: → I had Claude analyze a bunch of the lead editor’s past videos frame by frame, down to the tiniest detail, so every zoom, transition and sound matched the channel's established style. → I gave Claude very detailed instructions on editing a rough pass (dictated everything through Wispr) → Claude turned my notes into a build brief for Tesseract in Claude Code. → Tesseract built it in Premiere + After Effects → I'd analyze the output, and give Claude very detailed notes/instructions on what to fix or add. → Claude prepared a revision brief for Tesseract → Rinse and repeat until I was satisfied with the output To be clear, I still had to do a whole lot of manual editing. After 2 cycles of Claude not understanding what I was going for in certain segments, I just said "I'll do it myself" and did just that. all in all, it took 11 rounds of revisions to get it to a point where i was satisfied. And the most annoying part was that each round of revisions took like, an HOUR minimum. And I couldn't really do much on my computer in the meantime. BUT... On the bright side, the results truly were top tier. There were some sequences that would've taken me hours to produce in AE. But my agents fulled it off fast, fast! And the team loved the finished result which is the most important thing. I definitely learned a lot and I know my next attempt will be a lot faster. I have to remind myself that the people who get amazing results out of agentic editing are those who are building systems: skill files, style guides and checklists that get sharper with every project. And that’s what I'm doing. Everything this edit taught me is now written into a skill for the next one. It’s an upfront investment I'm prepared to make. I’ll report back on how much faster this next video goes (which I'm starting riiiiight NEOW!)