by PixVerseVideo
PixVerse
9 models · from 450 credits
Models in the PixVerse family
| Model | Mode | Price | Aspect ratios | |
|---|---|---|---|---|
| Pixverse c1 Image-to-Video | Image to video | from 650 credits | — | Open → |
| Pixverse c1 Reference-to-Video | Reference to video | from 650 credits | 16:9 · 9:16 · 1:1 · 4:3 · 3:4 · 2:3 | Open → |
| Pixverse c1 Start-End-to-Video | Start and end frame | from 650 credits | — | Open → |
| Pixverse c1 Text-to-Video | Text to video | from 650 credits | 16:9 · 9:16 · 1:1 · 4:3 · 3:4 · 2:3 | Open → |
| Pixverse v6 Image-to-Video | Image to video | from 600 credits | — | Open → |
| Pixverse v6 Reference-to-Video | Reference to video | from 600 credits | 16:9 · 9:16 · 1:1 · 4:3 · 3:4 · 2:3 | Open → |
| Pixverse v6 Start-End-to-Video | Start and end frame | from 600 credits | — | Open → |
| Pixverse v6 Text-to-Video | Text to video | from 600 credits | 16:9 · 9:16 · 1:1 · 4:3 · 3:4 · 2:3 | Open → |
| Pixverse v6 Video-Extend | Video extend | from 450 credits | — | Open → |
Prompting guide for PixVerse
TARGET MODEL: PixVerse (AISphere video). Sequential parser — earlier tokens weigh most. Priority order is official: subject identity → action → environment → technical cues LAST.
Rules that must hold
- Text-to-video is THREE sentences: S1 = subject + defining traits + ONE action + location. S2 = ONE camera move + named style/lens/lighting/composition cues. S3 = positive stability constraints. Sweet spot 50-80 words — past ~200 the prompt dilutes its own control.
- ONE action, restrained — competing verbs produce artifacts. Show speed as physical evidence (motion blur, streaking lights); never repeat "fast".
- ONE camera move per shot, named simply in prose ("slow macro push-in", "steady lateral tracking", "low-angle tilting upward").
- KILL the praise words — PixVerse's own guide says "cinematic", "beautiful", "epic", "professional" sample too broadly. Replace each with a named physical cue: "warm rim light", "cyan-magenta contrast", "35mm", "anamorphic 2.39:1", "one-point perspective".
- The prompt box takes POSITIVE constraints only: "Hands remain natural", "Cup shape remains stable", "Product silhouette stays intact". Naming a defect noun ("no bent fingers") can summon it — convert every exclusion to its positive form.
- Don't write "realistic" as a style word — realism is PixVerse's unstyled base; named cues carry the look.
- When the user wants sound, write the audio into the shot as content ("audio includes soft room tone and faint spoon clink") — never as a toggle or promise.
What works best
- Reference shape (official, verbatim rhythm): "A ceramic coffee cup sits on a dark wooden table as steam rises in slow curls. Slow macro push-in, warm tungsten side light, shallow depth of field, quiet morning cafe background. Cup shape remains stable, no text overlay, audio includes soft room tone and faint spoon clink."
- Character consistency is a vocabulary lock: keep the character's description a fixed paragraph (face, hair, outfit, build) and paste it WORD-FOR-WORD every shot — identical phrasing carries identity; drift in words is drift in face.
- Technical cues ride the tail, never the lead.
From LUVI's own manual for this family, the same rules the workspace rewriter and the MCP prompt engine apply.