FLUX 3 Video is here. Serious, fun, creative, real, cinematic, whatever you need it to be. Native audio, Text to Video, Image to Video with multiple frames, video continuation, dialogue in multiple languages. Comes with Draft mode so you can explore ideas fast at a fraction of Show more
视频模型
FLUX 3 文生 / 图生 / 首尾帧 / 延长视频生成器
在 Dreamega AI 上使用 Black Forest Labs 的多模态前沿模型 FLUX 3 生成视频。写提示词做文生视频，让参考图动起来做图生视频，用首帧和尾帧衔接过渡，或用视频延长续写已有的 MP4 片段。每种模式均可生成 5 至 20 秒、720p 或 1080p 视频，自带同步原生音频和多语言对白。
Flux 模型系列
FLUX 3 视频作品集
浏览用 FLUX 3 由提示词生成的片段。每个示例均为单次生成，自带同步音频。
AI Video
迟到五十年的外卖员
雷雨夜，一名骑车送餐的外卖员来到公寓门前，遇见了五十年后的自己，由文本生成。
Prompt
“A young bicycle courier arrives at an apartment door during a thunderstorm, but the digital display shows a date fifty years in the future. The scene opens with a fast tracking shot following him through wet neon streets, then becomes completely still as the door opens. An elderly man with the same face stands inside and looks at him in shock. The camera slowly rotates around both characters while lightning illuminates the corridor. The food bag begins glowing, and nearby objects float into the air. Emotional science-fiction mystery, realistic acting, controlled camera movement, cinematic neon lighting.”
Live PipelineTake 01 / 02
FLUX 3 图生视频作品集
看看 FLUX 3 如何让静态图片动起来。第一段片段由单张参考图生成；第二段则在首帧与尾帧之间，按提示词完成过渡。
Source Feeds01 Inputs
Pipeline
00%
Program · On AirAI · Generated
Output
Transcript · 01
废弃站台上的机器人
The old robot sits alone on an abandoned train platform, holding a faded ticket. Its head slowly rises as distant lights appear along the tracks. The station signs begin illuminating one by one, and a futuristic train silently enters the platform. The camera begins with a close-up of the ticket, tracks across the robot's weathered body, then cranes upward as the train arrives. Emotional science-fiction, quiet storytelling, atmospheric fog, warm and cold light contrast.
Live PipelineTake 01 / 01
FLUX 3 视频延长作品集
把已有片段接续成新的镜头。FLUX 3 从源视频结尾处继续，保持角色、光线与运动连贯。
Source Feeds01 Inputs
Pipeline
00%
Program · On AirAI · Generated
Output
Transcript · 01
博物馆骨架活了
Continue directly as the dinosaur skeleton begins moving. It carefully steps down from the exhibition platform, sending small bones and dust across the floor. The paleontologist runs through the gallery while other fossil displays begin moving behind glass cases. The camera tracks beside her, then swings around as she reaches the museum entrance and finds a living prehistoric jungle outside instead of the city. Maintain the same character, clothing, museum layout, lighting, and movement continuity.
FLUX 3 YouTube 视频
观看社区演示、评测与教程，了解 FLUX 3 的实际表现
- FLUX 3 Is Here — Everything You NEED to Know - ElevenLabs
- Introducing FLUX 3 - Black Forest Labs
- Black Forest Labs FLUX 3: First Look - Bryan Software
- 100+ FLUX 3 AI Videos Scarily Close to Real Life - Airt
- Flux 3 Review: Everything You Need to Know (2026) - Thrive Media
- FLUX 3 - Incredible AI Video Model - Demo Reel - Jerrod Lew
FLUX 3 YouTube 视频
观看社区演示、评测与教程，了解 FLUX 3 的实际表现
FLUX 3 在 X 上的热门评价
看看大家在 X (Twitter) 上对 FLUX 3 的评价
Introducing FLUX 3. One multi-modal model for Image, Video, Audio and Action-Prediction. Creations are truer to life in every kind of style. FLUX 3 Video is now available in early access (link below). Jointly trained in one unified architecture, our model can be extended to Show more
FLUX 3 Video is #2 in the world. To celebrate, FLUX 3 Video is free to use in our playground until Sunday 16th, 11:59pm PT (link in the thread). This is just the beginning of what’s coming. Up next: 4K, video editing, and multiple images & videos as reference input.
Last week, @bfl_ai launched their first video model, FLUX 3 Video. They have been testing an update on @arena, and the new version is ranked #2 in the Text-to-Video Arena! With 1496 pts, the updated FLUX 3 Video is just 16 pts behind the #1 spot, Gemini Omni Flash (1512 pts).
Action: An early version of FLUX 3 is now running on robots. @mimicrobotics was one of the first partners to gain early access to FLUX 3. Together we developed FLUX-mimic, a video-action model combining the FLUX 3 backbone with mimic's expertise in robot learning for dexterous Show more
Black Forest Labs just announced FLUX 3: a unified multimodal model for image, video, audio and action prediction. Founded in Freiburg, Germany, one of the few globally significant European companies in the AI sector. The current release is gated early access, not open source. Show more
FLUX 3 Video is here. Serious, fun, creative, real, cinematic, whatever you need it to be. Native audio, Text to Video, Image to Video with multiple frames, video continuation, dialogue in multiple languages. Comes with Draft mode so you can explore ideas fast at a fraction of
Black Forest Labs video model, which I posted about when it was first teased on their site back in August 2024, is about to arrive in the form of a multimodal Flux 3. A placeholder page was briefly up yesterday at bfl.ai/models/flux-3 before being quickly taken down. It Show more
Reel · Specifications
什么是 FLUX 3
Black Forest Labs 的多模态前沿模型，可生成自带原生音频的视频，已上线 Dreamega AI
- · 015-20s片段时长
- · 021080p最高分辨率
- · 034模式：文生 / 图生 / 首尾帧 / 延长
- · 047宽高比
FLUX 3 是 Black Forest Labs 的首个多模态前沿模型：它不再分别训练图像、视频和音频系统，而是在一个统一架构中联合学习这三者，同一主干网络还用于机器人动作预测。视频版本可生成最长 20 秒的片段，自带同步原生音频、带口型同步的多语言对白和电影级运镜控制。在 Dreamega AI 上提供四种模式：文生视频用提示词直接生成，图生视频让参考图动起来，首尾帧生视频衔接两张关键帧，视频延长则续写已有的 MP4 片段。每种模式均可生成 5 至 20 秒，支持 720p 或 1080p。
Reel · Capabilities
FLUX 3 的核心能力
Black Forest Labs 多模态模型在文生、图生与视频片段驱动生成上的表现
- Feature 01 / 08
统一的多模态模型
FLUX 3 在单一统一架构中同时学习图像、视频与音频，而非把独立系统拼接在一起，因此动作、声音与画面源自对场景的同一份理解。
- Feature 02 / 08
原生同步音频
对白、音效与环境音在生成视频的同一过程中产生，与画面中发生的事件紧密对应。音频默认开启，且不增加任何积分消耗。
- Feature 03 / 08
多语言对白与口型同步
角色可以用英语、中文、西班牙语、法语、德语、日语、葡萄牙语、俄语、意大利语、印地语等多种语言说话，口型动作与台词相匹配。
- Feature 04 / 08
单段最长 20 秒
时长可在 5 至 20 秒之间任选整数秒，整段片子一次生成完成，足以容纳多段落场景或一次完整运镜。
- Feature 05 / 08
首尾帧控制
上传首帧和尾帧图片，FLUX 3 会生成两者之间的过渡，并按照你的提示词呈现中间的动作、镜头运动与氛围。
- Feature 06 / 08
视频延长
将一段最长 15 秒的现有 MP4 片段续写为新的 5 至 20 秒镜头，同时保持相同的角色、服装、光线与动作连贯性。
- Feature 07 / 08
720p 与 1080p 输出
草稿和社交平台交付选 720p，成片选 1080p。各档均按秒计费，短测试成本低，只有成片才付全价。
- Feature 08 / 08
七种宽高比
可渲染 21:9、2:1、16:9、4:3、1:1、3:4 或 9:16，一次覆盖超宽银幕、宽屏、方形社交流与竖屏短视频，无需事后重新构图。
如何使用 FLUX 3 文生视频
用一段文字描述生成 5 至 20 秒带声音的片段
写提示词
描述场景、主体、运镜、光线和氛围。如果需要对白，把台词放在引号里，并注明语言。
FLUX.2 是 Black Forest Labs 的图像生成与编辑系列。FLUX 3 是全新的多模态前沿模型，在同一架构中联合学习图像、视频与音频，其首个发布版本是视频模型：可生成最长 20 秒的片段，自带原生同步音频与多语言对白。Dreamega 通过 WaveSpeed 提供该模型。
四种。文生视频用提示词直接生成片段；图生视频让你上传的参考图动起来，同时保持主体外观；首尾帧生视频生成首帧与尾帧之间的过渡；视频延长将一段最长 15 秒的现有 MP4 片段续写为新镜头。
输出为 720p 或 1080p。时长可取 5 至 20 秒之间的任意整数秒，整段片子一次生成完成。提供七种宽高比：21:9、2:1、16:9、4:3、1:1、3:4 和 9:16。
会。音频与画面同时产生且默认开启，涵盖对白、音效与环境音。对白支持多种语言，包括英语、中文、西班牙语、法语、德语、日语、葡萄牙语、俄语、意大利语和印地语，并带口型同步。如果只想要无声画面，也可以关闭音频。
视频延长接收单个 MP4 文件，最长 15 秒、最大 50 MB。新片段从输入结束处继续，因此请在提示词中描述接下来的动作，并将宽高比设置为与源视频一致。
所有模式均按生成视频的秒数计费：720p 为每秒 26 积分，1080p 为每秒 44 积分。5 秒 720p 片段为 130 积分，10 秒 1080p 片段为 440 积分。音频生成已包含在内，不额外收费。