Flux 3 v2v: Extend Any AI Video Clip Into a Full Scene
A 5-second AI-generated clip is a good starting point, not a finished deliverable. Whether the source comes from a text prompt, an image, or a reference video, that short clip often represents one beat in a longer intended sequence. Flux 3's video-to-video (v2v) mode is built for this: it takes an existing clip as input and generates a continuation or variation that matches the motion, style, and context of the original.
What v2v Does Differently
Text-to-video starts from nothing. Image-to-video starts from a still frame. Video-to-video starts from a clip — meaning the model has access to actual motion data: direction, speed, subject position across time, and environmental consistency from the source.
This makes v2v the right tool when:
- A t2v or i2v generation produced a result worth extending
- A scene needs a second shot that continues the action from the first
- B-roll needs to match the motion character of a primary clip
- The intention is to build a longer sequence from a controlled starting point
Flux 3 v2v is available at black-forest-labs/flux-3-v2v on Siray. It shares the same API structure as the flux-3-t2v and flux-3-i2v endpoints.
Parameters That Matter
Size: 720p or 1080p. Choose based on the resolution of the source clip — feeding a 720p source into a 1080p output is technically valid but the model is working with upscaled input data.
Duration: 5 to 20 seconds. The maximum 20-second output means a sequence of three v2v generations could build a full 60-second scene with consistent motion lineage.
Safety_tolerance: 0 to 4. Controls content moderation threshold. For general content, the default is appropriate. For platforms that allow more expressive content, adjusting this parameter unlocks the model's broader output range.
The prompt: Even in v2v mode, the text prompt shapes what the continuation does. A prompt that matches the source motion amplifies continuity. A prompt that introduces new direction can steer the clip away from its origin — useful for transitions, less useful for pure extensions.
Practical Use Cases
Scene extension: Generate a 10-second establishing shot with t2v, then use v2v to continue it into a close-up or action beat. The v2v output inherits the lighting, color grade, and physical movement from the source.
Continuity shots: When building a multi-clip sequence, feeding each output clip back as v2v input creates visual continuity across cuts that would otherwise require careful matching in post.
Style transfer at motion scale: The v2v mode can apply stylistic changes while preserving motion structure. A cinematic clip can become more documentary-style, or vice versa, without the motion reconstruction starting from scratch.
B-roll generation: Start from a primary clip featuring the subject, then generate supplementary angles or environment shots that match the scene's motion energy.
Developer Notes
Flux 3 is available on Siray at three endpoints:
- black-forest-labs/flux-3-t2v — text to video
- black-forest-labs/flux-3-i2v — image to video
- black-forest-labs/flux-3-v2v — video to video
Switching between modes is a single endpoint parameter change. The request format is normalized across all three, meaning a pipeline built for t2v can extend to v2v without restructuring the API call.
Pricing for t2v and i2v is confirmed at $0.260/second. The v2v rate has not been separately confirmed — check the model detail page in the Siray console for the current v2v pricing before building cost projections.
All three modes support 720p and 1080p output and duration values from 5 to 20 seconds.
Summary
Flux 3 v2v turns a short clip into a starting point for longer, directed scenes. The practical floor is a 5-second input producing up to 20 seconds of continuation — chain multiple passes and a single source clip can anchor a full sequence. The safety_tolerance parameter and text prompt give directional control on top of the motion inheritance from the source.
Create your free Siray account and start building video sequences with Flux 3 v2v.