Flux 3 Image to Video
Turn up to ten images into one FLUX 3 clip
- keyframes
- VIDEO
You've got the stills. FLUX 3 will turn them into the motion. This node is Black Forest Labs' hosted FLUX 3 model behind a ComfyUI input - you feed it 1 to 10 images and it animates them into a single video clip, with a soundtrack if you want one. It landed in ComfyUI core in August 2026, one of the first FLUX 3 video nodes to ship at all.
The classic use is storyboarding: generate the key frames locally with SDXL or Flux (your GPU, free), then let FLUX 3 breathe life between them. One image opens the clip; two morph from the first to the second; more get spread evenly across it. Because the heavy lifting happens on BFL's servers, your rig only uploads the images and downloads the result. No model files, no VRAM budget - just a Comfy account with credits.
How it works
The node packs your images and settings into a request, sends it to the flux-3-video endpoint on Comfy's API proxy, and polls until the clip is ready. It's a "partner node": billed per call through your Comfy account, priced by resolution and duration, so the graph shows a price badge before you run it. FLUX 3 also generates synchronized audio - ambient sound, speech, effects - unless you switch generate_audio off.
Your keyframes are validated before they go anywhere: each image needs at least 256 pixels per side and an aspect ratio no more extreme than 64:1, and you can't exceed 10.
The inputs that matter
keyframes- the images, in playback order. One opens the clip, two set start and end.placement- "spread across the clip" lets FLUX 3 place them; "at times" pins each to a second you choose (atimesfield appears, one comma-separated time per image, strictly increasing).duration- "auto" or a fixed length (5 to 20 seconds). Set it explicitly if you're spreading three or more images; the node errors otherwise, because spreading needs a known length.aspect_ratio,resolution- "auto" aspect picks from the prompt and inputs; resolution is 720p or 1080p.safety_tolerance- moderation strictness, 0 to 4. Requests carrying images are capped at 2 regardless, so don't waste time cranking it.seed- a lie in the nicest way: FLUX 3 picks its own seed, so results are nondeterministic no matter what you type. Bump it to force a re-run.
Output is a single VIDEO - wire it to a save/preview video node or right-click to save from the player.
Gotchas
Spreading multiple images with duration left on "auto" is the number one error - set the length. Watch the moderation: FLUX is BFL's censored line, and the strictness slider only goes so far when you send images. And remember this isn't a local node: if you lose internet mid-poll, the job may still complete server-side. When it works, this is the easiest way I've found to get a coherent multi-shot clip without stitching separate generations.
Inputs (9)
| Name | Type | Default | Description |
|---|---|---|---|
| prompt | STRING | How the scene should move and sound; the prompt is interpreted and expanded before generation. | |
| keyframes | COMFY_AUTOGROW_V3 | 1 to 10 images, in playback order. Minimum 256x256 pixels each. | |
| placement | COMBO | 'spread across the clip' lets FLUX 3 place the images (one opens the clip, two become its start and end); 'at times' pins every image to a second you choose. | |
| aspect_ratio | COMBO | auto | Output aspect ratio. 'auto' picks one from the prompt and inputs. |
| duration | COMBO | auto | Clip length in seconds. 'auto' fits the length to the content. |
| resolution | COMBO | 720p | Output resolution. |
| generate_audio | BOOLEAN | true | Generate synchronized audio (ambient, speech, effects). Off produces a video with no audio track. |
| safety_tolerance | INT | 20–4 | Moderation tolerance, 0 is the strictest. Requests that send images or video are capped at 2 whatever you set here. |
| seed | INT | 420–4294967295 | Seed to determine if node should re-run; FLUX 3 picks its own seed, so actual results are nondeterministic regardless of this value. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| VIDEO | VIDEO | — |