Blog
Model guides written from what we measured, honest comparisons, workflows that ship, and notes from inside LUVI.
Guides · 11Comparisons · 5Workflows · 2Inside LUVI · 4RSS feed
Featured · Inside LUVI
LUVI Creator is a web platform for generating video, images, audio and 3D with 223 AI models from 30 families, in one account paid in credits. It speeds work up with parallel runs, reusable references and workflows, and protects quality with prompting guides for 18 model families.
Guides
Hailuo takes its camera direction as a bracketed command from a closed list of fifteen, placed inline where the move happens—[Push in], [Tracking shot], [Static shot]. It is also picture only, so every word you spend on sound is wasted. Six seconds start at 380 credits.
Workflows
Two generations turn a script into a presenter: a voice model reads the text, then a lip-sync model puts that audio on a face. The voice step is cheap enough to redo a dozen times; the face step is not. Here is the order that saves the credits, with measured numbers.
Guides
Wan 3.0 reads an additive formula—entity, scene, motion, then the look, then the sound—and it renders audio in the same pass. Leave the camera unnamed and it will cut inside a clip you wanted as one take. Five seconds runs from 400 credits at 480p.
Comparisons
LUVI has six lip-sync families, and they do two different jobs: four animate a still portrait from an audio track, two re-drive the mouth in a video you already have. Which job you are doing also decides which clock you pay for—audio seconds or video seconds.
Guides
On Hunyuan3D's image-to-3D modes the text field is not read at all—the model's own contract says it accepts no prompt. Every bit of control you have is in how you prepare the input image. Meshes start at 600 credits on Rapid and 1,000 on Pro.
Inside LUVI
The feather button next to the prompt runs a per-family manual over what you wrote, for free. It has two layers: eight deterministic checks that only fire on things the manual can prove, and a dialect rewrite for the judgment. Here is what each layer can and cannot do.
Guides
Seedream 5.0 Pro has no aspect ratio parameter. The shape lives in size, written as WIDTH*HEIGHT with an asterisk, and a ratio typed into the caption is ignored. Here are the thirteen legal sizes, the thinking switch, and the caption it rewards. From 144 credits per image.
Comparisons
Four of the 224 models in LUVI's catalog expose a Negative Prompt field, and they are all Qwen Image. The field disappeared because the sampling method it depends on did. Here is what replaces it on each family, and the three places where writing "no X" still works.
Guides
A multi-model AI platform runs video, image, audio and 3D models from many labs in one workspace, on one balance. Choose one by the exact model versions it carries, whether it prices each run before it starts, and how it helps you prompt each model.
Comparisons
Higgsfield, Krea, OpenArt and LUVI Creator all run hosted MCP connectors that let Claude or ChatGPT generate on your account. They differ in whether you see the price before a run, how many models they expose, and whether the assistant gets each model's prompting guide.
Guides
Nano Banana Pro re-renders the entire image on every edit, so an instruction without a preservation clause changes things you never asked about. Say what changes, say what stays, and point at references as Image 1. Edits start at 175 credits at 1k.
Comparisons
Most video models on LUVI now render sound in the same pass as the picture, but they disagree about almost everything else: whether the switch is on, whether it costs more, and how you write the sound. Here is the audio switch and the audio syntax for each family, checked against their schemas.
Guides
HappyHorse generates picture and sound together, but it does not infer sound from what it sees: a prompt with no audio line returns a near-silent video. It is also one of the few video models that honors timecoded shots, which is the opposite of the Seedance 2.5 rule.
Guides
FLUX 3 Video renders picture and sound in one pass, so the audio is written into the prompt rather than switched on. Quote every line and give it a visible speaker, or it can appear as text burned into the frame. On LUVI, clips start at 1,870 credits for five seconds at 720p.
Guides
Kling 3 reads a shot as subject, movement and scene first, then camera, light and atmosphere: 60 to 100 words, one camera move, one action. The whole prompt is hard-capped at 2,500 characters and an overrun fails the submission before anything renders.
Guides
Seedance 2.5 reads integer-second timestamps, typed audio channels and one camera move per shot. Most bad results come from prompting it like Seedance 2.0. Here is the structure that works, taken from the guide we run inside LUVI.
Guides
Write Veo 3.1 prompts camera first: cinematography, subject, action, context, style. Put dialogue, effects and ambience into the shot, and switch on Audio Generation, which is off by default on LUVI. Veo 3.1 Lite has no audio. Clips start at 800 credits for eight seconds at 720p.
Comparisons
All three generate picture and sound in one pass, and all three read a prompt differently: H3 wants a fielded document, Omni wants a caption with a continuity clause, Wan wants an additive formula. Prices, limits and the one rule that breaks each of them.
Workflows
A phone photo of a bottle becomes a clean studio still with Nano Banana Pro, then a five-second clip with motion and sound on Seedance 2.5 — wired as one workflow, estimated before it runs, from about 3,200 credits.
Inside LUVI
LUVI shows a credit estimate before every generation and charges you when the result arrives. Since September the charge follows what the job really used — capped at the estimate, never above it. What changed, what we measured, and why one model dropped from 15 credits to 7.
Inside LUVI
iPhone HEIC photos used to fail in Chrome, WebP confused half the models, and one library's browser decoder took 27 seconds on a 12-megapixel shot. So we moved the conversion to the server and made one rule: every image is stored as JPEG or PNG, colour profile and orientation intact.
