Creating the Music Video 'Total Corruption' with MiniMax H3 Singularity and ComfyUI
User frankyboson used the MiniMax H3 Singularity model and ComfyUI workflow to create the first music video 'Total Corruption'.
User frankyboson used the MiniMax H3 Singularity model and ComfyUI workflow to create the first music video 'Total Corruption'.
Replicate releases a usage guide for Kling 4.0. The new version supports up to 30 seconds of generation, 720p/1080p output, four aspect ratios (including the new 21:9), and up to 10 keyframes. The keyframe feature can specify intermediate frames, allowing the video to pass through them sequentially in a single continuous shot. The prompt can be up to 26,000 characters.
Why it matters: Kling 4.0 adds continuous shot control with up to 10 keyframes, allowing creators to plan complex actions and transition rhythms accordingly.
The author set up a MiniMAX H3 i2v workflow via ComfyUI on a used RTX 2080 Ti (11GB VRAM). Due to a missing audio_vae resource, errors occurred repeatedly; after filling it in, video generation succeeded.
In a music video project, the author tested using a 2x2 storyboard grid generated with GPT image 2 as the reference image for MiniMax reference-to-video. The four panels, in reading order, correspond to the starting frame, two transition frames, and the ending frame. For complex narrative scenes, the adherence effect was better than using only prompts and first/last frames.
The key to turning a text script into an AI video is the workflow, not a single model: split the script by scene, generate a clip of about 4 to 15 seconds for each scene, then assemble them in order and add narration and music. VIBE, an iOS and Android app, integrates models such as Kling, Sora, and Veo, allowing users to generate scene by scene and export an edit on a phone.
The official Kling blog compared six CapCut alternatives—Kling AI, DaVinci Resolve, Clipchamp, Filmora, VEED, and VN Video Editor—across platforms, costs, and workflows. Kling 4.0 supports Omni Reference combining up to 15 reference assets, 3–30 second video generation, and up to 10 reference images as keyframes; Kling IMAGE 3.0 Omni can generate images up to 4K.
Kling 4.0 can use Omni Reference to combine up to 15 reference assets, using clear portraits, product images, and the like to guide the reconstruction of faces and details in blurry videos while preserving the original footage's motion, composition, and camera movement. In video editing scenarios, it supports up to 5 input clips (including 1 main video), with a combined duration cap of 30 seconds. The official guidance recommends first determining the cause of blur (motion blur, out of focus, low resolution, compression, low light), then deciding between sharpening, AI super-resolution, or reference-image reconstruction.
The official Kling blog released a guide to creating funny AI images, offering 12 creative directions and 6 customizable prompts based on Kling IMAGE 3.0 Omni, covering scenarios such as animals at work, anthropomorphic food, and historical figures traveling to the modern era. The model supports up to 10 reference images, native 2K/4K output, and can maintain consistent style across a series of images via Image Series.
The official Kling blog compared six beginner video editing software options: Kling AI, Clipchamp, iMovie, Filmora, CapCut, and DaVinci Resolve.
Canva released and launched its self-developed image generation model Canva Craft on October 22, using a two-stage structure that first plans the composition and then renders the image, supporting up to 2K resolution and more than 30 aspect ratios. The article proposes 8 items to verify before introducing it into production, including text accuracy, composition reproducibility, Magic Layers editing capability, consistency of people and small objects, stability of local edits, delivery dimensions, rights, and cost. The official documentation does not guarantee that a specific character will generate the same face every time, so actual output must be used to confirm.
The author tested Alibaba Qwen team's Qwen Image 2.1 in ComfyUI, using a single 7B model to perform text-to-image generation, multi-image editing with up to 10 reference images, and native RGBA transparent cutouts.
Why it matters: The author tested Qwen Image 2.1's multi-image editing and alpha channel output in ComfyUI, and provides directly reusable prompt formulas and parameter settings.
The author found that ChatGPT repeatedly failed to generate portraits of men with monolid eyes, because the conversational AI compressed the detailed "monolid eyes" condition emphasized by the user into the single word "monolid eyes," and also added unspecified clothing and performance details on its own. Switching to a model with higher reasoning ability improved the situation, but the image generation stage may still ignore this feature; the author recommends checking the instructions the conversational AI actually passes to the image generation AI.
Aimed at users making backgrounds with Midjourney and PixAI, this practical guide, based on official help documentation, organizes two paths for removing unwanted elements: adding --no people at the end of the Midjourney prompt before generation.
The author used Claude to build an automated workflow for ComfyUI, automatically queuing up 18 watercolor test images at 1:30 a.m. on a 16GB Mac, all of which were completed by 7:41 a.m. The specified auburn-haired character, however, appeared with white and silver hair in large numbers in the tag version, and this remained the case even after weakening the LoRA; the cause has not yet been confirmed. The author ultimately adopted the tag version, wrote auburn hair into the character setting sheet, and added white hair and silver hair to the negative items, to be verified in the next generation.
A ComfyUI beginner downloaded the MOSS-Image text-to-image workflow from Civitai, entered a prompt and ran it, but the queue shows 0 active and no image is ever produced. Their "Preview as Text (final prompt)" node is still empty after finishing, leading them to suspect the prompt is not reaching the sampler; the If/Else Switch's on_true and switch inputs and the loader node also appear to be unconnected.
A user on Reddit asked for help with a LoRA merging problem: they want to merge two LoRAs whose concepts and datasets are both different, hoping both can work well simultaneously and preserve their respective effects as much as possible, and asked which custom nodes, methods, and parameters are needed. The post gave no specific solution or result.
MOSS-Image in ComfyUI produces a 6-panel storyboard with Japanese descriptions and a table-based 絵コンテ in a single generation. All 18 Japanese descriptions across 3 イメージボード works rendered correctly, while 4 typos appeared across 2 pages of 絵コンテ. Single image generation takes approximately 18–40 seconds on an RTX 3060. The author provides region-separated Prompt writing techniques and a ComfyUI-MOSS-Storyboard custom node workflow.
Why it matters: Local MOSS-Image outputs storyboards with Japanese descriptions and table-based 絵コンテ in a single generation. The region-separated Prompt writing technique and ComfyUI workflow are directly reusable.
A user is asking about the best models, quantization, and workflow solutions for local Text-to-Video and Image-to-Video generation with ComfyUI on an RTX 5070 Ti (16GB GDDR7 VRAM) + Ryzen 5 7600X + 32GB DDR5 setup.
The author tested five local MiniMax H3 recipes using the same 15-second three-person alien village scene, with generation times ranging from 95 seconds to 843 seconds.
The author ran MiniMax H3 Singularity V2 Pruned INT8 ConvRot R2 with ComfyUI on an RTX 3060 12GB, generating an approximately 10-second video at 1536×864 resolution, 243 frames, and 4-step sampling, taking 608 seconds.
On the evening of October 25, the author used Meta's AI agent Muse to generate a 30-second vertical video for the self-made "Tianxia Buxian" series. The first three requests produced one 10-second trial and two 30-second versions; the last one was checked and found to be 720×1280, 24fps, 30.000 seconds, with an audio track.
The author tried out the ControlNet-LLLite for Anima in ComfyUI, which can use 3D poses to specify character actions and region control to specify the generation range.
A user is trying to generate loop videos in Minimax using the same first-and-last-frame method, but the motion always "starts" at the 1st second and decelerates in the last 1 second, making it impossible to achieve a seamless constant-speed transition. Taking a 15-second hammering loop as an example, the actual result is 1 second of startup, 13 seconds of steady motion, and 1 second of wind-down before jumping back to the beginning. The user has tried prompts such as "video is a clip of the action segment" with no success, and is now seeking a working workflow, LoRA, or prompt solution.
A father plans to use ComfyUI on Runpod to make a short film in which his daughter, who loves Game of Thrones, marries her partner in a medieval castle surrounded by family. He says he is new to image generation and workflows, is learning by trial and error, and expects to need DaVinci for editing later, and he is publicly asking for advice.
AIwood Lab found during H3 VBVR testing that loading VBVR did not meet expectations and was even worse; after switching to the older ComfyUI 0.38 environment and comparing with 0.42, noticeable differences appeared in scenes such as the chef.
On a machine with an RTX 3060 Ti (8GB VRAM) and 16GB of RAM, the author used Qwen-Image-2.1 to generate an image of an underwater train carriage, then used MiniMax H3 together with Veda and Endless to extend it segment by segment into a 25-second video.
Before the PR video is finished, watch the exported file in 3 passes: first watch the content from a first-time viewer's perspective, then look at the text and the lower part of the frame, and finally listen to the audio only. Classify the problems you find as "unclear communication," "product distortion," or "unreadable/inaudible," and note the seconds and change points in a revision note. Change only one place at a time; after each change, bump the version number, re-export, and keep the old version. When done, archive the video together with the work description note, labeled "Fictional case · self-produced."
The author used Animagine XL 4.0 Opt in ComfyUI to generate a Halloween little devil with silver hair, red eyes, short black horns, wearing a black-and-orange outfit and looking up at the camera; the resulting image is 1280×1024.
Creator Naoya disclosed the workflow for making the 18-minute 24-second, 4K, 129-shot Sekigahara-themed historical film "Shimazu's Retreat" using tools such as MiniMax H3. The warlords' lines in the film were switched to Japanese voice acting; the author wrote the lines in hiragana without kanji, wrapped them in dedicated tags, and added an instruction to close the mouth after speaking, to reduce misreading and text appearing in the frame. After watching all 129 shots one by one, they were reshot, fixing issues such as helmets disappearing, two people speaking at the same time, reversed forms of address, and backgrounds that looked like a modern park; the ending was changed to a 34-second fade-out of the theme song.
The author built a MiniMax H3 workflow in ComfyUI, combining SelfLift progressive resolution sampling with a DLSS 5 upscaling node to output 2K or 4K video on an RTX 3060 with 6GB of VRAM. In their 6GB VRAM environment, generating and upscaling to 4K took about 35 minutes, with parameters of 8 steps and 0.6 megapixels, all without additional model training; both the workflow and the DLSS5 DLL download links have been released.
Why it matters: The author tested the full pipeline from low-resolution progressive sampling to 4K upscaling on a 6GB VRAM graphics card and provided the time taken and parameters, as a reference for users with low-spec devices.
Pollo AI's Recast feature can reference the original video's motion and camera movement while replacing the characters and background; the author used it to convert a live-action two-person dance into original Japanese-style animation characters. The same material was generated at 480p and 720p respectively; 480p consumed 235 credits and 720p consumed 505 credits, a difference of 270 credits. The 720p details are slightly clearer, but the 480p result already exceeded expectations, and the author recommends trying 480p first.
Claude set up a ComfyUI workflow on an M1/16GB Mac; after adjusting launch options, single-step generation went from 10 minutes to about 10 seconds, and they combined processes including a watercolor LoRA, black-and-white manga switching, inpainting, and upscaling.
The author used Animagine XL 4.0 Opt in ComfyUI to generate an example of a female rock singer holding a mic, leaning forward, and belting, at a resolution of 1248×832, Seed 2026101857, Steps 28, CFG 5, euler_ancestral. The accompanying JSON uses only 7 standard nodes and requires no input image, LoRA, ControlNet, or custom nodes; two comparison JSONs are also provided, one removing the forward lean and one removing the singing expression.
A user tested a high token limit on MOSS SFT with a prompt of about 1.3k tokens, combined with a 4-steps turbo LoRA, 4MP resolution, and Euler/simple sampling, to generate a photo containing 12 different people. Each character's appearance, clothing, and props remained consistent, with no color bleeding or duplication.
In a local MiniMax H3 image-to-video workflow, MOSS-Image (5B) native 4MP output combined with a simple 3–4 word prompt generated a roughly 3-second clip, with detail retention and temporal consistency exceeding expectations. The suede jacket texture, skin pores, chains, tweed and thin frames all showed no texture sliding, and the second video also preserved the stylized elongated neck proportions rather than forcibly correcting them. The test was based on an NVIDIA RTX PRO 6000, with the built-in MOSS template in ComfyUI as the workflow.
The author shares hands-on results from training a character LoRA on MOSS Image, saying that using Automagic3 with Sigmoid and the High setting, a learning rate starting at 0.0005 with a minimum learning rate set, and not excluding block 22 produced good results every time.
Aizawa Amu posted 5 illustrations on X from the morning of October 24 to late night on October 25, and freely distributed the corresponding ChatGPT image-generation prompts, covering five scenes: morning, daytime, evening, night, and late night. The prompts require attaching a reference image of an adult character aged 21 or over that the user has the right to use, preserving facial features, hairstyle, hair color, and eye color, and outputting a 1536:1024 landscape single-scene anime-style illustration. The author also suggests changing only one detail, for example swapping the daytime plaid fabric sample for a polka-dot pattern.
A user trained a style LoRA for Qwen-image with ai-toolkit on a Colab 80GB A100, but the output backgrounds were too simple, random people and duplicate characters appeared, and the style drifted. The configuration was rank 16, 1024px, bf16, adamw8bit; a learning rate of 1e-4 for 500 steps had no effect, so it was changed to 3e-4 and trained to 2k steps, and a 12-image dataset still showed no improvement. The user plans to expand to 30–40 images.
Using Animagine XL 4.0 Opt with Control-LoRA Canny rank128 contour guidance, a 1248×832 landscape composition of an elf drawing a bow with Seed 2026101813 was generated in ComfyUI, and the contour PNG and a JSON containing the node configuration were released.
A collection of 100 AI beauty image-generation prompts for tools such as ChatGPT and Gemini has been released, all in directly copyable JSON format. Each prompt set includes prompt, negative_prompt, and aspect_ratio fields, and provides 1:1, 9:16, and 1280×670 sizes plus background and mood adaptation versions. The themes focus mainly on beauties in spectacular and everyday scenes around the world, such as Argentine salt flats and Kyoto riverside terraces.