FLUX 3 is a multimodal video model from Black Forest Labs that can generate shots with optional synchronized sound from text, image keyframes, or existing clips. This platform uses flux-3 and currently supports text and image generation, video continuation, and draft enhancement. It is suitable for previewing ideas first, then advancing selected options into full-quality short films.
Submit requests to the public API at api.acedata.cloud using the documented parameters, then use the results in your application.
api.acedata.cloudflux-3Input parameters and result formats vary by service. Use the public API for this model and follow its guide for generation, task retrieval and editing operations.
The above reflects the current generation options available on this platform. Refer to the generation modes and parameters listed on this page.
First prepare images with clear subjects and connectable viewpoints; keyframes can arrange frames in sequence or by time. Suitable for product appearances, lighting changes, and simple transitions; images constrain the starting and ending states, while the model still generates the motion in between.
v2v uses start_video to provide an existing clip, then describes the action that happens next. It is suitable for developing a usable shot into the next scene; check the subject, direction of motion, and visual continuity before adding it to the edit.
HD drafts can be used to validate composition, shots, and pacing. After selecting a direction, use your own draft_task_id to request full quality while retaining the original prompt and assets; caching is temporary, so enhancement should not be treated as a permanently reproducible capability.
Use product keyframes to define the appearance, then specify presentation actions, camera angles, and ambient sound in the text. Check product details and on-screen text first, then use the finished video as advertising or social content material.
Write subject actions, dialogue, sound effects, and ambient atmosphere into the same shot brief; sound can be generated with the visuals. Review dialogue, lip sync, and audio timing segment by segment, and revise in post-production when necessary.
Use drafts to compare ideas, keyframes to define transitions, and continuation to develop selected shots. Check continuity after each segment is generated; the narrative and pacing of a complete work still need to be organized through an application or editing workflow.
If you already have product images, storyboard stills, or shots that can be continued, FLUX 3 provides a clear input path. It can start from text and continue creating with keyframes and existing clips, rather than offering only single-image first-frame animation.
Drafts are available only in HD. Before increasing the output tier, confirm the direction of the action, composition, and sound; higher clarity does not automatically fix dialogue, subject shape, or narrative errors.
Use t2v to write a complete shot brief, or use i2v and keyframes to constrain the visuals; use v2v and start_video to continue an existing clip.
Set action=generate, model=flux-3, and mode for /flux/videos; standard generation is 5–20 seconds, and continuation is 5–15 seconds. Drafts are HD only; save draft_task_id.
Select draft_enhance for draft enhancement, use your own valid draft_task_id, and retain the original prompt and assets; after async, retrieve the final video and media information through /flux/tasks or a callback.
A product image defines a small radio on a table by the window, with its knob slowly turning, the camera gently pushing in, accompanied by clear knob sounds and quiet room ambience.
First confirm the direction with an HD draft and retain your own draft_task_id; when enhancing, do not change the original prompt or assets, and check the visuals and sound at full quality.
You can currently use action=generate, with mode used internally to select t2v, i2v, v2v, or draft_enhance.
Select mode=i2v and provide keyframes and a shot prompt. Keyframes can be an image list or image pairs arranged by seconds; first check image perspective and action continuity.
Use start_video with v2v to provide an existing clip, then describe the subsequent shots with a prompt.
Generation mode supports optional synchronized sound; describe dialogue, sound effects, and ambient sound in the prompt. generate_audio=false can request a silent video. You should still listen to and check the dialogue after the video is completed.
Save the draft_task_id returned by the draft, then select draft_enhance while the cache is valid and ownership is the same. Enhancement reuses the original prompt and assets, so you cannot change the creative content at the same time.
Use async=true to first obtain task_id, then query through /flux/tasks or use callback_url; wait for success, then save the video link and actual media information.
Choose a development approach or application suited to this model and follow its setup guide.
↗Python SDK Integration Guide
SDK installation, authentication and examples↗TypeScript SDK Integration Guide
SDK installation, authentication and examples
↗This is only a basic call example. See the full docs for more parameters and advanced usage.
View Docs