Pixio's first model Montage: prompt-to-film, 7-minute shorts with per-scene retakes
tsi_org · x · 2026-11-15
Pixio launched Montage, its first model and the engine behind Director: give it a pitch or a song and it writes the treatment, casts people and sets, shoots every section and cuts the finished film.
Key specs:
- Music videos planned against your song, lip-synced in 15s sections up to 6 minutes
- Short films run 30s–7min, from an idea or your own clip prompts
- Draft-first workflow: retake any single section, upscale to 1080p, grade, or face-fix per section
- 16:9 / 9:16 / 1:1, up to 8 parallel productions per account
- Multimodal I/O: text, up to 9 reference images, audio (new songs written from lyrics), up to 3 reference clips
Related event: Pixio Launches Montage, Generating Up to 7-Minute Videos From One Prompt(3 posts)→
More from Multimodal
- mark_k: Killing Grok Imagine would be a huge mistake, ceding image/video to Chinese models — mark_k · 2026-11-15
- Omni Video Generation Space Trends on Hugging Face — CumaSystem · 2026-11-15
- AI glitches become a Seinfeld script: a hobbyist builds his own ComfyUI front-end — AxonkaiLab · 2026-11-15
- A 5,000-word Claude prompt wrangles 9 reference characters into a 20-second Subaru ad — Nimblecloud13 · 2026-11-15
- UChicago study: artists can engineer visual memorability, but computational models capture it better — plopesresearch · 2026-11-15
- Creator finds MiniMax H3 base R2V beats bf variants for voice cloning — Foreforks · 2026-11-15