Pushing H3 text-to-video with layered prose prompts that Qwen3VL actually understands
SIR_NVAX_A_LOT · reddit · 2026-09-06
A Redditor shares an H3 text-to-video experiment stress-testing how well Qwen3VL handles non-standard, prose-heavy prompts — with surprisingly good comprehension.
Key techniques: a layering approach composing the shot through long, cascading prose that defines the subject in detail (face material, luminous hair filaments, carved star-chart wardrobe, lighting); int8 quantization at 20 steps, 720p; quick test-runs at 2s/int8/8 steps. The prose style adds randomness, so composition and lighting sometimes took a third attempt. The final piece stitches four clips together, with the score also generated by H3 using a 32x32 canvas trick. A full example prompt (an extremely detailed layered description of an "artificial divinity" character) is included.
More from Multimodal
- Two Midjourney sref codes for grainy pastel fashion illustration style — sergeantsref · 2026-10-23
- Three free ways to try MOSS-Image: API playground, HF Space, ComfyUI — songguoxiansen · 2026-10-23
- Open-source MOSS-Image tops 1M ComfyUI downloads, 5B model rivals closed rivals — songguoxiansen · 2026-10-23
- GAFU Uses Generation-Side Representations to Boost Fine-Grained Visual Understanding in Unified Multimodal Models — Tengju Ru · 2026-10-23
- MiniMax Design MCP goes live: create with Design inside Codex and other agents — Hailuo_AI · 2026-10-23
- Quiver's detailed SVG generation sparks debate: code-drawn images could replace pixel-based image models — stuffyokodraws · 2026-10-23