40 Imagen 4 Fast drafts, then a Nano Banana 2.1 edit: $1.10
Draft 40 text-to-image options on Imagen 4 Fast at $0.025 ($1.00), then refine the pick with one Nano Banana 2.1 edit at $0.10. Total $1.10 on Sume.

A cheap two-stage flow on Sume is to draft 40 options on Imagen 4 Fast, pick one, and refine it with a single Nano Banana 2.1 edit. The cost is 40 x $0.025 = $1.00 for the drafts plus $0.10 for the edit at 1K, so $1.10 in total.
How the hand-off works
Imagen 4 Fast is text-to-image only on Sume, so it cannot refine its own output. Sume returns Sume-hosted URLs for results, and Nano Banana 2.1 accepts public HTTPS reference URLs, up to 10. Pass the chosen draft URL in input_references with a prompt that states the fix: correct the logo, change the wall color, remove the cup. Use the Sume media URL that the job returns.
| Stage | Model id | Count | Arithmetic | Cost |
|---|---|---|---|---|
| Draft | google/imagen-4-fast | 40 | 40 x $0.025 | $1.00 |
| Refine at 1K | google/nano-banana-2.1 | 1 | 1 x $0.10 | $0.10 |
| Total | 41 | $1.10 |
Limits to plan for
Imagen 4 lists five aspect ratios: 1:1, 16:9, 9:16, 4:3 and 3:4. If the final needs 4:5, 4:1 or 8:1, draft in a close ratio and let the Nano Banana 2.1 pass set the final one with its own ratio list, which includes 4:5, 4:1 and 8:1. Note that Google's Imagen page says Imagen is shut down in the Gemini API; check that the Sume row still lists, or swap the draft stage to another cheap row.
When not to bother
If the final will be a photo edit anyway, skip the draft stage and send the real photo to Nano Banana 2.1. The two-stage flow is for ideas from a blank page.
Cost control
Imagen 4 Fast is the cheapest row in the set, so 40 drafts cost the same as 10 images on Nano Banana 2.1 at 1K ($1.00 either way). That is the argument for the flow: for a dollar you see 40 ideas rather than 10.
If you want twice the ideas, 80 drafts cost $2.00. The refine step stays at one call. The refine is the part that depends on the model: at 2K the edit costs $0.15 and at 4K it costs $0.20, so a 4K refine after 40 drafts totals $1.20.
Log the prompt of the winning draft. It is the best prompt you have and the refine prompt should reuse its subject words.
Quality of the hand-off
Check the picked draft at full size before you spend on the refine. If the composition is wrong, the edit will not fix it. If only details are wrong, the edit is the right tool: a wrong logo, an extra object or a color shift.
Also note the cost of rejected refines. If the first edit fails, a second one costs another $0.10. Even with three attempts the flow totals $1.30, the same as 13 Nano Banana 2.1 images at 1K.
For a blog or social team, the same flow scales to a weekly budget. Ten posts a week at $1.10 each is $11.00 a week, or about $47.63 over a 4.33-week month (11 x 4.33). Compare that with generating each post's image once at Nano Banana 2.1 1K for $0.10: $4.33 a month. The flow costs more because it buys choice, so use it where choice matters, such as ads, covers and hero art, and skip it for filler images.
Sources
Related posts
More in Use cases
- Four-language product video from recorded voiceovers: $1.20
Reuse one edit for English, Korean, Spanish and French. Swap the audio spine per language: four Sume renders and four caption jobs cost $1.20.
- Google Ads Shorts previews: render 9:16, 16:9 and 1:1 first
Shorts previews need a Video view or reach campaign with Shorts selected. Render all three Google sizes from one master with Sume Timeline for $0.30.
- Google Ads video rejects MP3, WAV and PCM: pair audio with a still
Google Ads video upload does not accept audio files such as MP3, WAV or PCM. Turn a radio spot into an MP4 with a still and your audio using Sume Timeline.
- Indie game trailer: one 30 s Seedance 2.5 job, 480p then 720p
A 30-second game trailer on Seedance 2.5 is $8.07 at 480p, $17.34 at 720p and $42.65 at 1080p on Sume. Draft cheap, pay once for the final.
Written by Sume