Extract Frame
Pull a single still frame out of a video as a PNG image.
Overview
The Extract Frame node converts a frame of video into a still image you can route into image-based nodes (modify, edit, upscale, generate-image as reference, etc.). Six pick modes:
- First — the first frame of the video
- Last — the final rendered frame
- Timestamp — a specific second-precision offset (e.g., 2.5s into the clip)
- Nearest keyframe — snap to the nearest keyframe at or after a given timestamp (faster, no inter-frame decode; defaults to
0= first keyframe) - Frame # from start — extract by frame index (
0= first frame); worker probes source fps to seek precisely - Frame # from end — extract by frame index from the end (
0= last,1= second-to-last, etc.); worker probes duration + fps
Configuration
| Field | Type | Default | Description |
|---|---|---|---|
| Mode | Select | first | One of the six modes above |
| Timestamp (s) | Number | 0 | Seconds into the video (used by timestamp and keyframe) |
| Frame # from start | Number | 0 | Frame index from the start (used by frame-index) |
| Frames back from end | Number | 0 | Frame index from the end, 0 = last (used by frame-from-end) |
Inputs & Outputs
Inputs: Video (required, connected upstream or via Upload Video) Outputs: Image (PNG)
Credit Cost
1 credit per extraction, regardless of mode or video length.
Best Practices
- Use last mode to grab the closing frame of a generated video for use as the start frame of a follow-up i2v generation — that’s the trick behind continuous multi-clip storytelling.
- Use first mode to verify what an upstream video starts with when chaining nodes.
- Use frame # from end when you want “the second-to-last frame” or “10 frames back from the end” without caring about the source’s exact length.
- Use frame # from start when you need an exact frame index (e.g., frame 24 for the second of a 24fps clip).
- Use nearest keyframe when extraction speed matters or when the source is high-bitrate — keyframe extraction skips inter-frame decoding (
-skip_frame nokey). - Timestamp mode honors decimals (e.g.,
2.5for 2.5 seconds in). Out-of-range timestamps clamp to the nearest valid frame. - The output image inherits the video’s resolution — to resize it before downstream use, route through Resize Video upstream or use a generative node’s resolution control.
Common Use Cases
- Capture the final frame of a VEO/Kling clip to seed the next i2v step (perfect-loop or storyboard-chain workflows).
- Pull a representative still for a thumbnail before posting to social.
- Extract a specific moment from a long video for editing or reference.
- Grab the start frame of an upstream clip to use as a reference image for a Generate Image node.
- Extract “the frame just before the end” (e.g., 5 frames back) when the last frame has dissolve artifacts —
frame-from-endwithframesFromEnd=5.
Tips
- Frame extraction does NOT incur Topaz / VEO / KIE provider charges — it’s a local FFmpeg operation costing one Nodaro credit for orchestration.
- The frame is exported as PNG (lossless), not JPEG — safe to route into upscale or edit-image without compression artifacts.
frame-indexandframe-from-endhard-fail if the source’s fps or frame-count probe yields an invalid value (rather than silently mis-converting).- Combine with Combine Videos to take a still from clip A, run it through Generate Image to remix, and stitch the result back into a new clip.