4K output at $0.1/mp/s — broadcast/campaign video workflows could wrap here
FLUX Upscale: Standalone Video Upscaling to 2K and 4K Now Available
Original: FLUX Upscale: 2K and 4K for Video
Importance: FLUX 3 Video既存ユーザーに対する実用的な機能拡張だが、直接的な破壊的変更や全体的なビジネス影響は限定的
Summary
Black Forest Labs has released FLUX Upscale as a standalone tool and API endpoint. It regenerates generated videos from 704p to native 4K resolution, preserving identity and reference details better than general upscalers. Two modes are offered: Precise (4 steps, $0.07/mp/s) for identity consistency and Creative (8 steps, $0.1/mp/s) for enhanced detail generation. Supports 1.5x, 2x, and 3x upscale factors. Designed for broadcast, campaign, and big-screen formats that require higher pixel density. Available via BFL API and playground.
Key Points
- Upscale video from 704p to native 4K resolution
- Two modes: Precise for identity consistency, Creative for detail enhancement
- Optimized for FLUX 3 output with superior detail preservation
- Targets broadcast, campaign, and large-format delivery
- Available immediately via API endpoint and playground
View developer notes (APIs, breaking changes, migration)
FLUX Upscale deployed as dedicated endpoint 'flux-3-upscale'. Natively understands FLUX 3 output and repairs details general upscalers miss. API parameter upscale_factors supports 1.5x, 2x, 3x (704p input → ~1080p, 2K, 4K). Precise mode (4 steps, $0.07/mp/s) prioritizes identity/reference consistency; Creative mode (8 steps, $0.1/mp/s) enhances repair and detail generation. Available at BFL dashboard playground (model=flux-3-upscale) and docs.bfl.ai/flux_tools/flux_video_upscale. Underlying tech already powers 1080p step in FLUX 3 Video.
Source: https://bfl.ai/blog/flux-video-upscale
Outlet: Black Forest Labs
This article is an AI-generated summary (OpenAI GPT-4o-mini) of publicly available information from Anthropic, OpenAI, Google, Meta, Mistral, DeepSeek, Sakana, and other vendors. The original source URL is always provided in accordance with fair-use citation requirements. Summaries are AI-generated and may contain mistranslations or misinterpretations. Always verify details with the original source.