Six Nano Banana 2.1 start frames, six Wan 3.0 clips: a 3-min Short
Six 9:16 Nano Banana 2.1 stills ($0.60 at 1K) feed six 30 s Wan 3.0 jobs as first frames: $23.40 at 720p for a 180-second Short, with Timeline.

Generate six vertical stills first, then pass each one as the first frame of a 30-second Wan 3.0 job. At 1K the stills are $0.60 ($0.10 each); six 30-second 720p clips are $22.50; the Timeline join is $0.30. Total: $23.40 for a 180-second Short, the maximum length (YouTube Help, read 2026-10-09).
The reason to start from stills is control. Nano Banana 2.1 went GA on 2026-10-06 with better multi-turn character consistency and new aspect ratios (Gemini API changelog, read 2026-10-09), so you can lock a character in the image step and let the video step animate it.
Budget
Sume's image catalog lists 9:16 in the aspect_ratio values and 1K or 2K as resolution tiers. Wan 3.0 accepts 2 to 30 seconds and first_frame / last_frame images.
| Stills | Video | Stills cost | Video cost | Timeline | Total |
|---|---|---|---|---|---|
| 1K ($0.10 each) | 720p | $0.60 | $22.50 | $0.30 | $23.40 |
| 2K ($0.15 each) | 720p | $0.90 | $22.50 | $0.30 | $23.70 |
| 1K ($0.10 each) | 1080p | $0.60 | $45.00 | $0.30 | $45.90 |
The two requests
Step one asks for a vertical still. Step two passes the finished image URL as frame_images. When both frame_images and input_references are sent, frame_images decides and the job is image-to-video.
# 1) the still
curl -X POST https://api.sume.com/v1/images \
-H "Authorization: Bearer $SUME_API_KEY" -H "Content-Type: application/json" \
-d '{"model":"nano-banana-2.1","prompt":"Scene 3: ...","aspect_ratio":"9:16","resolution":"1K"}'
# 2) the clip, first frame = that image
curl -X POST https://api.sume.com/v1/videos \
-H "Authorization: Bearer $SUME_API_KEY" -H "Content-Type: application/json" \
-d '{"model":"wan-3.0","prompt":"Slow push-in, scene 3","aspect_ratio":"9:16","resolution":"720p","duration":30,
"frame_images":[{"type":"image_url","image_url":{"url":"https://example.com/scene3.png"},"frame_type":"first_frame"}]}'Gotchas
- Use the same character description in all six image prompts; the model does not remember the previous call unless you pass a reference.
- Each frame_images entry needs a frame_type of first_frame or last_frame.
- Import the six finished MP4s before the Timeline render; Timeline only accepts media.sume.com URLs from your workspace.
- Stills are billed even if you discard the video, so check each still before you submit the clip.
Sources
Related posts
More in Use cases
- Six podcast ad-read variants at 500 characters: TTS cost and tags
Six 500-character ad reads cost $0.1425 on Sume TTS ($0.02375 each). Tag each take with the metadata field and one Idempotency-Key per variant.
- A 60-second Short as twenty 3-second Omni cuts: $7.50 at 720p
Twenty 3-second Omni Flash 1.1 clips at 720p cost $7.50 and at 1080p $11.25, the same as six 10 s clips. Twenty slots pass 12, so Timeline chunks the render.
- Spelling bee: 200 word-pronunciation clips on Sume TTS, one job each
200 words of about 12 characters each cost $0.1140 on Sume TTS as 200 jobs, $0.000570 each. One 2,400-character job is $0.1140 but gives one file.
- Split a 14-minute episode into 20 clips with overlaps: 2 cents
Detach the audio once ($0.01), then one timeline audio split with 20 overlapping ranges ($0.01). A 860-second track fits the 900-second detach cap.
Written by Sume