3090+3060 local video gen: which model — Wan 2.2, MiniMax H3 or LTX-2?
itspacedev · reddit · 2026-10-09
A user built a local video-generation server (RTX 3090 24GB + 3060 12GB, Xeon E5, 32GB RAM, Linux with ComfyUI and wan2gp) and wants realistic 5–10s 720p text-to-video and image-to-video, but generation is very slow. Earlier community advice: Wan 2.2 works on a 3090 but 720p takes 15–20 min per clip; MiniMax H3 int8 with a 4-step LoRA was suggested as an alternative.
Open questions: which model + LoRA combo yields 5s 720p in 5 min with acceptable quality on a 3090; whether the 3060 is useful for text encoding or upscaling; whether 32GB RAM suffices; and any good ComfyUI workflow for submitting jobs from a script.
More from Infra
- GPUs already within 2x of brain efficiency, and still beat human workers on energy — MikePFrank · 2026-10-09
- Do AI agents still need Kubernetes? Berlin event says yes, with agent-on-K8s cases — Al_Grigor · 2026-10-09
- Cloud Backlogs Hit $1.69T, CoreWeave Posts $2.58B Quarter as Inference Becomes the Battleground — FinanceYF5 · 2026-10-09
- NVIDIA Is AI's Central Bank: A100 Paper Citations Still Beat H100+H200 Combined — FinanceYF5 · 2026-10-09
- Browser-Based Calculator Crunches the Real Power Cost of Self-Hosted LLMs vs Cloud APIs — paq85 · 2026-10-09
- PartyKit shuts down free hosted platform 2.5 years after Cloudflare acquisition — threepointone · 2026-10-09