H3 video performance transfer: full prompt makes AI characters act convincingly
R34vspec · reddit · 2026-10-06
A detailed Reddit writeup demonstrating performance transfer with H3 video generation, using the same seed as a prior test. The author shares the full prompt, whose core is retentionanalysis (declaring how much of the reference video's visuals/audio to preserve) plus subjectdefinitions pulling appearance from image 1, voice timbre from audio 1, scene from image 2, and outfit from image 3. A detaileddescription blocks out the shot—rainy rooftop, shallow depth of field, slow push-in—with line-by-line dialogue and micro-expressions, plus accent and voice specs. The result: noticeably better emotional acting from the generated character, and the prompt is reusable for anyone wanting more convincing character performances.
More from Multimodal
- MIRRORSIDE: a behind-the-scenes film set that never existed, made with AI — StrategyMedium5907 · 2026-10-06
- Qwen-Image 2.1 edits look unfinished: user shares ComfyUI params seeking fixes — Suspicious_Aide2697 · 2026-10-06
- Creator hands repetitive workflow to Codex and GPT-6 Astra, keeps creative calls — socialwithaayan · 2026-10-06
- Midjourney style code sref 705714994 yields whimsical brush-texture characters — michaelrabone · 2026-10-06
- AI-Created Crowd Looks Indistinguishable From Real People, Creator Warns — minchoi · 2026-10-06
- Runway CEO invokes Jevons paradox: coding agents are expanding software demand, video is next — c_valenzuelab · 2026-10-06