Carousel covers that move
- Reels
- 10
- Clients
- 8
- Made in Flow
- 10
- Credits, approximate
- 304
- Made with
- Credits
- Agent time, minutes
Prompt used
Slide 1 is the one that moves
Google Flow, step by step
- Model used
- Omni 1.2 Flash
- One 10 s take
- 30 credits
≈ US$0.60 allocated - This run, two takes
- 60 credits
≈ US$1.20 allocated - Native MP4
- 720 × 1280
24 fps · 10 seconds
About this run
The slide goes into Google Flow as both the first and the last frame, so the clip ends where it began and loops.
Run on 18 Nov 2026 with a slide that had not been animated before. 60 credits, about three minutes of generating.
Flow: cost and output quality
- Cost: about US$0.60 for one 10-second take, or US$1.20 for this two-take run, allocating a US$19.99/month Google AI Pro seat across its 1,000 included credits. These are allocated subscription costs, not separate per-video invoices; unused credits raise the effective cost per delivered reel.
- Measured output: 720 × 1280 at 24 fps. The middle 720 × 900 was enlarged 1.5 times to the delivered 1080 × 1350. Exporting at 1080 does not create native 1080 detail.
- Observed quality: the chosen take retains the layout and readable lettering, with slight softness from generation and enlargement. Its motion has more frames than the 6.7 fps GIF from this test.
- Two different results: take 1 had cloud shadows and loose leaves over the lettering. Take 2 kept the words clear, with faint wisps beside the green pill. Only one of the two takes was selected, but both used credits.
- Evidence: the settings screenshot below names the model and shows the 60-credit cost. Both takes and lettering close-ups are shown on this page. Google's model credit rates · Google AI Pro pricing. Checked 18 Nov 2026.
Start with the exported slide Straight from Photoshop or the AI output
a- aSlide 1, nothing changed
Pad it to 9:16 Flow only makes 16:9 or 9:16
a
bc- aOriginal slide
- bSame slide, centred, not resized
- c285 px bands in the slide's edge colours
Start a new project flow.google.com, on a Pro account
a
- aNew project
Set it, check the cost Flow shows the credits before you generate
a
b
c
d
e
f
- aVideo, Frames
- b9:16
- cOmni 1.2 Flash
- d720p, 10 s
- eTwo takes
- f60 credits
Upload the padded frame Click Start, upload, then pick it. Repeat for End
a
b
c
- aUpload media
- bPick the padded file, not the raw slide
- cAdd to prompt
Write the prompt, generate Same frame at both ends, so it loops
a
b
c
d
- aStart frame
- bEnd frame, same image
- cPrompt: lock the camera, name what moves
- dGenerate
Two takes come back About three minutes
a
b
c
d
- aTake 1
- bTake 2
- cPadded frame
- dRaw slide
Download each take The 720p original, not the 1080p upscale
a
b
c
- aMore options
- bDownload
- c720p, original size
Crop back to 4:5 Cut off the same bands that were added
ab
c- aFlow's clip
- bKeep the middle 4:5
- cFinished reel
Check every take Same prompt, two different results
a
Take 1b
Take 2- aShadows and loose leaves cross the type
- bType untouched, motion stays on the ground
- Flow redraws typeLay the original text back on top
- Skip the 1080p upscaleIt adds texture. Upscale the 720p file instead
- Check every takeIt can invent faces, leaves or objects
ChatGPT GIF, step by step
Same slide, same day, in ChatGPT on the CCM workspace. No padding step.
- GIF engine name
- Not shown
- Per-GIF price
- Not shown
- Usage source
- Included usage
- Native GIF
- 276 × 346
6.7 fps · 16 frames
ChatGPT: model, cost and output quality
- Model identity: this conversation does not display the underlying GIF generation model. We cannot reliably attach a GPT-Image model name to this output.
- Cost: the workspace usage panel shows included usage. No separate per-GIF charge is displayed, and this run's dollar cost and exact quota consumption are unavailable. It should not be described as free; subscription and workspace limits apply. OpenAI's ChatGPT pricing documentation.
- Measured output: 276 × 346, 16 frames, 2.4 seconds, about 6.7 fps. We repeated it four times to 9.6 seconds and enlarged it about 3.9 times for the 1080 × 1350 MP4.
- Observed quality: the lettering and illustration are visibly softer after enlargement, with fewer animation frames than the Flow take. The slide is redrawn at a small size, rather than preserving its original full-resolution artwork. Nothing crossed the lettering in this run.
- Scope: these measurements describe this downloaded GIF, not a maximum resolution or a universal quality rating for ChatGPT. The prompts differed between methods, so this is an output comparison rather than a controlled model benchmark.
Start with the exported slide The same 1080 × 1350 file
a- aSlide 1, nothing changed
Attach the slide New chat, plus button, Add photos and files
a
b
- aPlus button
- bAdd photos and files
Ask for a looping GIF Say what stays still and what moves
a
b
c
- aAttached slide
- bPrompt
- cSend
One GIF comes back In the chat, already looping
a
b
c
d
- aYour slide
- bYour prompt
- cThe GIF, click to open
- dPlayback speed
Open it and download Download arrow, top right
a
b
c
- aSheet or GIF view
- bDownload GIF
- cThe 16 frames
What the file is Shown to scale against the slide
a
- aThe slide
- bThe GIF: 16 frames, 2.4 s, 0.8 MB
Convert to MP4 Enlarge 3.9 times and repeat the loop

a
1080 × 1350 MP4, 9.6 s- aLooped four times, enlarged to the slide's size
The two methods, side by side
Same slide, one run each. The prompts were not identical: the ChatGPT one asked for less motion.




| Measured on this run | Google Flow | ChatGPT GIF |
|---|---|---|
| Generation model | Omni 1.2 Flash | Underlying GIF engine not displayed |
| What you upload | Slide padded to 9:16 first | Slide as exported |
| File that comes out | 720 × 1280MP4 | 276 × 346GIF |
| Enlargement to 1080 × 1350 | 1.5 timesthen crop | 3.9 times |
| Length | 10 s | 2.4 srepeated to 9.6 s |
| Frames a second | 24 | 6.7 |
| Takes per run | 2 | 1 |
| Cost per take | 30 creditsfor 10 s, about US$0.60 allocated | Per-GIF price not displayed |
| Cost of this run | 60 creditsabout US$1.20 allocated on Pro | Included workspace usage; dollar cost unavailable |
| Observed detail | Readable type; slight softening after 1.5× enlargement | Softer type and illustration after 3.9× enlargement |
| Layout | Original layout retained in the selected take | Slide redrawn at a small size |
| What went wrong | Take 1: shadows and leaves over the type | Nothing crossed the type |
| Extra step to MP4 | Crop | Enlarge, loop, convert |
What it costs
One seat makes 16 reels a month
Omni 1.2 Flash, 720p, 10 s, two takes. Google AI Pro includes 1,000 credits.
01,000 credits
- One 10 s take
- 30 credits
- One reel
- 60 credits
- Per reel
- $1.20
- Per seat
- $19.99
What these versions used
Displayed versions only; earlier attempts excluded.
RecordedEstimated
Nine designers, one seat each
$179.91a month for the team, whatever the volume up to 144 reels.
| Each designer makes | Reels a month | Seat used | Cost per reel |
|---|---|---|---|
| 1 reel a week | 36 | 240 of 1,000 | $5.00 |
| 2 reels a week | 72 | 480 of 1,000 | $2.50 |
| 4 reels a week | 144 | 960 of 1,000 | $1.25 |
How these numbers were worked out
- Flow charges 3 credits a second for Omni Flash 1.2 at 720p, so a 10-second take is 30 credits. Two takes per reel is our usual setting.
- $19.99 for 1,000 credits is 2 cents a credit, so 60 credits is $1.20 when a seat is fully used.
- The table counts four weeks a month and no retries. A retry adds 60 credits.
- Monthly credits do not roll over. Extra credits can be bought on Pro; Google does not list pack prices.
- Nine designers is the current weekly design team. Each needs their own Google account.
- The four stronger-motion redos on 18 Nov used 30 credits each: Due Season, Samaritan Health, Life Springs and Faith Fellowship carousel 1 (120 total). Earlier attempts are excluded from the bars. Mauldin used 60, Godchasers 24 and Faith Fellowship carousel 2 used 10. Cathedral and both Covenant reels are estimated at 30 each.
- Agent time for the unchanged reels comes from session logs. The four redos were not individually timed.
- The Pro account held 348 credits after the four redos on 18 Nov 2026 and resets on Dec 13.
- US list prices from Google, checked 18 Nov 2026: plans and Flow credits. Local pricing, taxes and promotions may differ.
Other Flow models: published cost, quality not tested
| Model / setting | One generation | Credits / allocated USD | Quality evidence here |
|---|---|---|---|
| Omni Flash 1.2 · 720p | 10 s | 30 · ≈ $0.60 | Tested above: 24 fps, readable type in the selected take |
| Omni Flash 1.2 · 360p | 10 s | 10 · ≈ $0.20 | Not tested |
| Veo 3.1 Lite · 720p | 8 s | 10 · ≈ $0.20 | Not tested |
| Veo 3.1 Fast · 720p | 8 s | 20 · ≈ $0.40 | Not tested |
| Veo 3.1 Quality · 720p | 8 s | 100 · ≈ $2.00 | Not tested |
| Omni Flash 1.1 · 720p | 10 s | 15 · ≈ $0.30 | Not tested in this comparison |
- Rates shown for Google AI Pro / non-Ultra users, one generation each. Two takes double the credits.
- USD figures allocate the US$19.99 Pro seat across 1,000 fully used credits; they are not standalone clip prices.
- Google lists Veo 3.1 Fast, Veo 3.1 Quality and Omni Flash 1.1 for retirement on 5 Dec 2026.
- A model's name or higher price is not a measured quality result. Only Omni 1.2 Flash at 720p was verified for this walkthrough; the ChatGPT GIF engine was not disclosed.
- Published Google credit rates and retirement dates, checked 18 Nov 2026.