Introduction

PixVerse has a new flagship. PixVerse V6 Max is now the top model in PixVerse's V series, a closed text-to-video and image-to-video model that makes up to 15 seconds of 1080p video with synchronized audio. It's a closed model, and you can run it today on PixVerse's official Replicate page, on WaveSpeed (live since October 21), and inside ComfyUI, where a new partner node landed on October 22.

Think of it as V6 with the quality dial turned all the way up: same inputs, but PixVerse says it follows prompts better, renders skin, fabric, and light more naturally, and keeps characters and products consistent from shot to shot.

What V6 Max is good at

From PixVerse's model readme on Replicate:

  • Long, detailed prompts: up to 5,000 characters for text-to-video, enough to write a full storyboard with camera moves and performance notes
  • Emotional acting: eyelines, breathing, pauses, and body language, not just facial expressions
  • Consistency across shots: characters and objects keep their identity as angles and scenes change
  • Music and rhythm: movement and transitions are built to land on the beat
  • Multi-character dialogue: lines go to the right speaker, lip sync is tighter, and characters who aren't talking keep their mouths shut
  • Products and short text: packaging, colors, and short brand copy stay legible through close-ups and rotations

Specs and switches

  • Modes: text-to-video and image-to-video (image-to-video follows your image's aspect ratio)
  • Resolution: 360p, 540p, 720p, or 1080p
  • Duration: up to 15 seconds (Replicate offers 5, 8, 10, or 15; WaveSpeed lets you pick anywhere from 1 to 15)
  • Audio: one switch adds background music, sound effects, and dialogue; put spoken lines in quotes and name the speaker
  • Multi-clip: another switch turns a "Shot 1, Shot 2, Shot 3" prompt into a cut sequence
  • Also: negative prompt and seed for repeatable tests
  • Not supported: first/last-frame transitions, so stick with regular PixVerse V6 for those

Where to run it and what it costs

Prices are per second of output and change with resolution and audio, and they differ a lot by host:

  • Replicate (PixVerse's official listing): $0.11/s at 720p silent up to $0.28/s at 1080p with audio. A 5-second 720p clip with audio is $0.75.
  • WaveSpeed: $0.049/s at 720p silent up to $0.124/s at 1080p with audio
  • ComfyUI partner node: choose "PixVerse V6 Max" from the model menu on the PixVerse V6 partner node (it is merged on the main branch, so grab a fresh nightly or wait for the next tagged release). The price badge shows a launch rate of about $0.146/s at 1080p with audio until November 3, then about $0.178/s.

Same model, very different bills, so price-check before you queue a batch of 15-second takes.

A prompt to try

This one uses multi-clip and audio together. Turn on both switches, choose 1080p, 15 seconds, 16:9:

Shot 1, wide shot at golden dusk: a lighthouse keeper in her sixties with a silver braid and a navy wool sweater climbs a spiral iron staircase, lantern in hand, waves crashing far below, slow upward crane move. Shot 2, medium close-up: she reaches the lamp room, sets the lantern down, and says softly, "Right on time, old friend," as she wipes salt from the glass; warm gold light on her face, aqua sea glow through the windows. Shot 3, close-up of the great lens beginning to turn, light beams sweeping across violet storm clouds, camera slowly pulls back through the window to reveal the whole lighthouse against the sea. Sound: wind, distant surf, creaking metal steps, a low cello melody that swells as the beam sweeps out. Photorealistic, filmic color grade, natural textures, steady cinematic camera.

Original Source

PixVerse's official model page and readme: pixverse/pixverse-v6-max on Replicate.

Day-0 API listing: PixVerse V6 Max on WaveSpeed.

ComfyUI partner node: Comfy-Org/ComfyUI pull request 17287.

Conclusion

V6 Max is PixVerse aiming straight at short dramas, ads, and music videos: long storyboard prompts, cut-together shots, and dialogue with sound in a single render. PixVerse hasn't posted independent benchmarks yet, so treat the quality claims as the vendor's until you've run your own side-by-side, but the multi-clip and audio switches make it a fun one to test this weekend.

—Aurelia ♡