---
title: Run a model shoot-out for your studio
display: 'Run a *model* shoot-out for your studio'
nav_title: Model shoot-out
summary:
  - "A dated scorecard that names your team's model for one recurring job, with a runner-up, cost and wait."
  - "Change only the model: same prompt, same reference, same aspect ratio, each at its highest quality setting."
  - "Four runs, one per model, then a second seed for the top two; started in parallel, they finish in the time of the slowest."
description: "Pick the model for a recurring job with a fair, repeatable shoot-out: same input, same prompt, a scorecard, and a recorded winner and runner-up."
section: playbooks
order: 410
type: playbook
topic: Studio and production
level: Build
roles: [team-lead, artist]
time: { read: 9, do: 25, wait: 4 }
needs: ["Any plan with credits", "One real brief from your team's work", "The reference images that brief uses"]
last_verified: 2026-10-16
models_checked: ["GPT Image 2.5 Sunburst", "GPT Image 2.5 Flare", "Nano Banana 2.1", "Seedream 5.0 Pro"]
hero: playbooks/compare-models-for-your-studio/aurelle-shootout-four-models.webp
related: []
run_in: [model_openai-gpt-image-2-5-sunburst, model_openai-gpt-image-2-5-flare, model_google-gemini-nano-banana-2-1, model_bytedance-seedream-5-0-pro]
# the use case made, on scenario.com/explorations, and the platform workflows that run it
made_with:
  - { slug: nine-models-one-brief, title: "Nine Models, One Brief" }
thumb: playbooks/compare-models-for-your-studio/thumb.webp
video: tutorials/compare-models-for-your-studio.mp4
chapters:
  - { at: 0, title: "Four models, one brief", image: "/media/images/playbooks/compare-models-for-your-studio/ch-01-four-models-one-brief.webp" }
  - { at: 18.5, title: "A shortlist from the ranking", image: "/media/images/playbooks/compare-models-for-your-studio/ch-02-a-shortlist-from-the-ranking.webp" }
  - { at: 47.5, title: "The product keep-list, then the scene", image: "/media/images/playbooks/compare-models-for-your-studio/ch-03-the-product-keep-list-then-the-scene.webp" }
  - { at: 56, title: "Change one thing: the model", image: "/media/images/playbooks/compare-models-for-your-studio/ch-04-change-one-thing-the-model.webp" }
  - { at: 77, title: "Same size, same crops", image: "/media/images/playbooks/compare-models-for-your-studio/ch-05-same-size-same-crops.webp" }
  - { at: 101.5, title: "A second seed for the top two", image: "/media/images/playbooks/compare-models-for-your-studio/ch-06-a-second-seed-for-the-top-two.webp" }
  - { at: 114, title: "Winner and runner-up", image: "/media/images/playbooks/compare-models-for-your-studio/ch-07-winner-and-runner-up.webp" }
  - { at: 125.5, title: "Project Instructions, read-only", image: "/media/images/playbooks/compare-models-for-your-studio/ch-08-project-instructions-read-only.webp" }
  - { at: 137, title: "Recap", image: "/media/images/playbooks/compare-models-for-your-studio/ch-09-recap.webp" }
steps:
  - title: "A shortlist from the ranking"
    text: "Create › Images › View all models, with Input Image and Output Image ticked and the list ranked by Recommended. The run took the top two, GPT Image 2.5 Sunburst and Flare, and added Nano Banana 2.1 and Seedream 5.0 Pro; the input is the approved studio image of the car."
  - title: "The product keep-list, then the scene"
    text: "GPT Image 2.5 Sunburst, the studio image in Reference Images, Dimensions 1536x1024, Quality high, Image Count 1. Write the test prompt once: the product's keep-list, then the test scene."
    prompt: |-
      The exact car from the reference image, with its body shape and proportions, liquid satin champagne paint, one continuous thin LED light bar across the full width of the nose, the spaced AURELLE wordmark in chrome capitals centered under the light bar, wide dark lower intake with fine horizontal slats, dark twisted multi-spoke turbine wheels with gold brake calipers, and black glass roof unchanged, parked on a gravel overlook on a high mountain pass at sunrise. Snow-capped peaks and a winding road behind, low warm sun from the left raking along the side of the car, cold blue shadows, thin mist in the valley below. Three-quarter front view at headlight height, the light bar lit and the AURELLE wordmark facing the camera and sharp. Soft contact shadow under the tires on the gravel. 3:2 landscape. No other cars, no people, no other text or logos. Style: automotive launch photography, controlled reflections, perfect panel gaps, no other brands.
  - title: "Change one thing: the model"
    text: "Keep the prompt and reference and switch to GPT Image 2.5 Flare, then Seedream 5.0 Pro (2048 x 1360), then Nano Banana 2.1 (Aspect Ratio 3:2, Resolution 2K). Never tune the prompt per model, or you compare prompts, not models."
  - title: "Same size, same crops"
    text: "Line the four up at the same size and crop the front of each at the same scale. All four kept the wordmark and light bar, so paint and light decided: Nano Banana went cool and silvery, Seedream looked overcast and was the slowest."
  - title: "A second seed for the top two"
    text: "Sunburst and Flare were close, so each ran once more with the same inputs. Both held the car again and the ranking didn't change."
  - title: "Winner and runner-up"
    text: "Score wordmark, light bar, wheels and calipers, paint, light, artifacts and wait: Sunburst wins for its light, Flare is the runner-up for drafts. Write it in Settings › Project › Agent › Project Instructions, and keep the prompt and reference as a test pack to rerun when a new model ships."
agent_steps:
  - title: "Connect the Scenario MCP"
    text: "Connect your agent to the Scenario MCP and sign in once at [mcp.scenario.com](https://mcp.scenario.com). Then tell it which Scenario team and project to use for the session, so every output lands there."
  - title: "Give the agent the files"
    text: "The approved studio image of the car, the reference for every model. Upload it with the agent or give it its asset ID. Use a real brief from your team's work, not a showcase prompt."
  - title: "Brief the agent"
    text: "One prompt, one reference, one take per model: only the model changes. The agent can also start from recommend for image to image at quality priority to build the shortlist."
    prompt: |-
      Run a model shoot-out on one brief. Change only the model: the same prompt, word for word, and the same reference for every run. Never rewrite the prompt for a model.

      Reference: the studio image of the car, passed as referenceImages, an array with its asset ID.

      Prompt, word for word:
      The exact car from the reference image, with its body shape and proportions, liquid satin champagne paint, one continuous thin LED light bar across the full width of the nose, the spaced AURELLE wordmark in chrome capitals centered under the light bar, wide dark lower intake with fine horizontal slats, dark twisted multi-spoke turbine wheels with gold brake calipers, and black glass roof unchanged, parked on a gravel overlook on a high mountain pass at sunrise. Snow-capped peaks and a winding road behind, low warm sun from the left raking along the side of the car, cold blue shadows, thin mist in the valley below. Three-quarter front view at headlight height, the light bar lit and the AURELLE wordmark facing the camera and sharp. Soft contact shadow under the tires on the gravel. 3:2 landscape. No other cars, no people, no other text or logos. Style: automotive launch photography, controlled reflections, perfect panel gaps, no other brands.

      Candidates and settings, one image each, at each model's highest quality:
      - GPT Image 2.5 Sunburst (model_openai-gpt-image-2-5-sunburst): width 1536, height 1024, quality high
      - GPT Image 2.5 Flare (model_openai-gpt-image-2-5-flare): width 1536, height 1024, quality high
      - Nano Banana 2.1 (model_google-gemini-nano-banana-2-1): aspect ratio 3:2, resolution 2K
      - Seedream 5.0 Pro (model_bytedance-seedream-5-0-pro): 2048 x 1360

      Read each model's parameters with model_schema_get first; where a model offers only aspect presets, pick 3:2. Quote the cost with a dry run first and wait for my go.
      Launch the four with wait: false and collect them with one jobs_wait. Report each job's cost, wait and output size.
  - title: "Line them up and score"
    text: "The agent can lay out the grid and fill the cost, wait and size columns from the jobs; the visual scores stay yours. A wordmark error eliminates a model however good the rest looks. In this run all four held the wordmark and light bar, so paint and light decided."
    prompt: |-
      Download the four results, put them in one grid at the same size, and crop the front of each (light bar, wordmark, intake) at the same scale into a second grid. Write a scorecard with one column per model and these rows: wordmark spelled correctly, light bar, wheels and calipers, paint vs reference, light and mood vs brief, artifacts, cost, wait. Fill cost and wait from the jobs and leave the visual rows for me.
  - title: "A second seed for the top two"
    text: "When two models are close, run each once more with the same inputs before you decide. Here Sunburst and Flare held the car again and the ranking did not change."
    prompt: |-
      Run GPT Image 2.5 Sunburst and GPT Image 2.5 Flare once more each, with exactly the same prompt, reference and settings. Show both new takes next to their first takes.
  - title: "File the test pack and report"
    text: "Record the winner and runner-up with the date, and keep the inputs to rerun when a new model ships."
    prompt: |-
      Create a collection called "Model test pack: product scenes". Add the studio reference and every take to it, and put the test prompt and the settings of each model in its description. Report every asset ID with its model, the cost and wait of each job, and the total cost spent.
updated: 2026-10-16
canonical: https://learn.scenario.com/playbooks/compare-models-for-your-studio
---

Last updated: 2026-10-16

# Run a model shoot-out for your studio

> Pick the model for a recurring job with a fair, repeatable shoot-out: same input, same prompt, a scorecard, and a recorded winner and runner-up.

In short:

- A dated scorecard that names your team's model for one recurring job, with a runner-up, cost and wait.
- Change only the model: same prompt, same reference, same aspect ratio, each at its highest quality setting.
- Four runs, one per model, then a second seed for the top two; started in parallel, they finish in the time of the slowest.


A scorecard that names the model your team uses for one recurring job, with a runner-up, the cost and wait of each, and the inputs to rerun it when a new model ships. For team leads and leads of craft.

:::note As of October 2026
The models, settings and screens in this article are the ones Scenario offered in October 2026. New models and updates arrive often, so check the app for newer options before you choose.
:::

![Four images of the same champagne grand tourer on a gravel overlook on a mountain pass at sunrise, one per model](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/2f8933b245fc598cf9555113ca8462389e4e60c6e72a8132f94785cd180b10a2.webp "Same prompt, same reference. Top left: GPT Image 2.5 Sunburst. Top right: GPT Image 2.5 Flare. Bottom left: Nano Banana 2.1. Bottom right: Seedream 5.0 Pro.")

## The brief

| Field | Brief |
|---|---|
| Job | Product in a location scene from a studio reference |
| Input | The studio launch image of the Aurelle Solenne, an invented electric grand tourer in liquid satin champagne |
| Test | The car on a gravel overlook on a mountain pass at sunrise, three-quarter front, 3:2 |
| Candidates | Four image models that take a reference, picked from the platform ranking for image to image at quality priority |
| Fixed | Prompt, reference, aspect ratio, one output each, quality high |
| Score | Wordmark, light bar, wheels and calipers, paint vs reference, light and mood vs brief, artifacts, price per accepted image, wait |
| Decide | Winner and runner-up, recorded with the date |

The job tested is the one the [Product lifestyle scenes](https://learn.scenario.com/playbooks/product-lifestyle-scenes.md) playbook does all day: put a branded product from a reference into a new scene.

![The Aurelle Solenne in a dark studio: champagne paint, a thin light bar across the nose, the AURELLE wordmark under it, dark turbine wheels with gold calipers](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/3788fd18dedcb41fcc5001c006e6f41df8394fea7ccc5cfe18f57eb4d5f5aa02.webp "The reference: the approved studio image.")

## Approach

:::flow
Shortlist | Platform ranking, image to image, quality |
Test prompt | Product keep-list plus test scene | text
Run four models | Sunburst, Flare, Nano Banana 2.1, Seedream 5.0 Pro | image
Line up | Same-size grid and front crops |
Score and record | Winner, runner-up, cost, wait, date |
:::

A fair shoot-out changes one thing: the model ([Compare models on your brief](https://learn.scenario.com/courses/pick-the-right-model)). The prompt stays word for word, with the same reference, the same aspect ratio and the highest quality setting each model offers.

Use a real brief from your team: a model that wins on a generic landscape can lose on your paint color or your wordmark. Shortlist with data, then decide with eyes:

1. **Shortlist** three or four candidates from the capability ranking. In this run the platform's recommendation for image to image at quality priority ranked GPT Image 2.5 Sunburst, GPT Image 2.5 Flare, GPT Image 2, Meta Muse Image Edit and Seedream 5.0 Pro; GPT Image 2 was left out as the older sibling of the first two, and Nano Banana 2.1 was added as a strong model from another family. Use **Choose a Model** with the **Image to Image** capability and **Quality: Highest First**, or ask Scenario Genie ([Choose a model](https://learn.scenario.com/courses/pick-the-right-model)).
2. **Run** each once with identical inputs. One take per model is enough to eliminate; run a second seed for the top two when they are close.
3. **Score** on the criteria the brief cares about, in order. A wordmark error eliminates regardless of beauty.
4. **Record** the winner and runner-up with cost, wait and date, and keep the inputs to rerun later.

## Steps

:::steps
1. **Write the test prompt once.** It is the product keep-list plus the test scene:

   ```prompt
   The exact car from the reference image, with its body shape and
   proportions, liquid satin champagne paint, one continuous thin LED light
   bar across the full width of the nose, the spaced AURELLE wordmark in
   chrome capitals centered under the light bar, wide dark lower intake with
   fine horizontal slats, dark twisted multi-spoke turbine wheels with gold
   brake calipers, and black glass roof unchanged, parked on a gravel
   overlook on a high mountain pass at sunrise. Snow-capped peaks and a
   winding road behind, low warm sun from the left raking along the side of
   the car, cold blue shadows, thin mist in the valley below. Three-quarter
   front view at headlight height, the light bar lit and the AURELLE
   wordmark facing the camera and sharp. Soft contact shadow under the tires
   on the gravel. 3:2 landscape. No other cars, no people, no other text or
   logos. Style: automotive launch photography, controlled reflections,
   perfect panel gaps, no other brands.
   ```

2. **Run model 1.** **Create** > **Images**, **GPT Image 2.5 Sunburst**, the studio image in **Reference Images**, **Dimensions** 1536x1024, **Quality** `high`, **Image Count** 1.

3. **Run models 2 to 4** ([Run a same-input shoot-out](https://learn.scenario.com/browse)) with the same prompt and reference: **GPT Image 2.5 Flare** at the same settings; **Nano Banana 2.1** with aspect ratio `3:2` and resolution `2K`; **Seedream 5.0 Pro** at 2048x1360. Where a model offers only aspect presets, pick the matching ratio.

4. **Line them up.** Put the four results in one grid at the same size, and crop the front of each (light bar, wordmark, intake) at the same scale into one image.

   ![Front crops from the four models in a grid, each showing the light bar and the AURELLE wordmark](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/158924edb00c5c06e75ee8c0276c42939a11d8bd84d31fe844b31d9b161e1367.webp "Top left: Sunburst. Top right: Flare. Bottom left: Nano Banana 2.1. Bottom right: Seedream 5.0 Pro. All four spelled AURELLE and kept the light bar; the decision moves to paint and light.")

5. **Score each criterion** and fill the scorecard below. Look at the reference while scoring the product.

6. **Run a second seed for the two closest.** Sunburst and Flare were close, so each ran once more with the same inputs. Both held the car again; the ranking did not change.

7. **Record the decision** in the team's model notes and set the winner as the default in the team App or workflow for this job.
:::

**With the Scenario MCP:** call `recommend` with the capability (here image to image) and quality priority to get the shortlist, `model_schema_get` for each candidate's parameters, then four `model_run` calls with `wait: false` and one `jobs_wait`.

An agent can write the scorecard skeleton (cost, wait, output size) from the job results; the visual scores stay human.

## Check the result

:::check
**Accept the comparison when:** every model received the same prompt and reference; outputs are compared at the same size; the wordmark and the product were checked at 1:1 against the reference; cost and wait come from the actual jobs.
**Reject the comparison when:** prompts were tuned per model; one model ran at a lower quality setting; the winner was picked from thumbnails; the brief was a showcase prompt instead of real work.
:::

The scorecard from this run (1 = poor, 5 = excellent):

| Criterion | Sunburst | Flare | Nano Banana 2.1 | Seedream 5.0 Pro |
|---|---|---|---|---|
| Wordmark spelled correctly | yes | yes | yes | yes |
| Light bar, turbine wheels, gold calipers | held | held | held | held |
| Paint vs reference (liquid satin champagne) | 5 | 5 | 3 (cooler, silvery) | 3 (greyer) |
| Light and mood vs brief (sunrise, raking light, mist) | 5 | 4 | 3 (flat light, car smaller in frame) | 3 (cool, overcast light) |
| Artifacts | none seen | none seen | none seen | none seen |
| Price per accepted image, read from the quote | same as Flare and Nano Banana 2.1 | same | same | highest of the four |
| Wait | about 40 s | about 27 s | about 35 s | about 126 s |

- **Winner: GPT Image 2.5 Sunburst**, for the light that best matched the brief: low sun raking along the flank, mist in the valley, controlled reflections on the paint.
- **Runner-up: GPT Image 2.5 Flare**, which held the product on both seeds at about two thirds of the wait and is the better pick for drafts and volume.
- **Nano Banana 2.1** had the largest native output (2528 px wide at 2K) but turned the champagne cooler and silvery, which matters for a paint color.
- **Seedream 5.0 Pro** held the car's shape but read as an overcast morning rather than sunrise, and was the slowest and most expensive of the four.

One run per model only screens the candidates: the second seed for the top two is what made the winner a decision rather than a lucky take.

:::compare
![GPT Image 2.5 Sunburst take: the champagne car on the gravel overlook, low sun raking along its side, mist in the valley](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/24f2b7e473eb64d20b8bd004186531b6aa0ba8dcfb8100f67a705294d54f62f1.webp "Winner: Sunburst. Champagne paint and raking sunrise light closest to the brief.")
![Nano Banana 2.1 take: the car smaller in frame on a gravel overlook in flatter light, the paint cooler](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/f4a08e2c5b22cdec7a4bf8628ff7fcd717c79fce7ee19de20ff8f707e17bfe96.webp "Nano Banana 2.1: wordmark right, paint turned cooler and silvery, light flat.")
:::

:::compare
![Sunburst second seed: the same car and overlook, framed slightly wider, sun low on the left](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/aad0166cade48c60c52e1be2c2a368e19fca4797196ac5b74b14411c9171bb6b.webp "Sunburst, second seed: the product held again.")
![Flare second seed: the same car and overlook, mist in the valley, sun on the left](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/fe74ee1f336c37409bfaf76de056b865aa87ff7b7e82512f2d51c249508ff797.webp "Flare, second seed: the product held again; close to Sunburst.")
:::

:::quiz Check yourself
? Same prompt and studio reference, two models. The brief asks for sunrise light raking along the car and mist in the valley, and the paint must match the reference. Both wordmarks read correctly. Which take wins?
![The champagne car on a gravel overlook below snowy peaks, in even morning light](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/f4a08e2c5b22cdec7a4bf8628ff7fcd717c79fce7ee19de20ff8f707e17bfe96.webp "Take A")
![The champagne car on a gravel overlook below snowy peaks, the sun low on the left](https://ufgc0c9mhl4ueq76.public.blob.vercel-storage.com/academy/24f2b7e473eb64d20b8bd004186531b6aa0ba8dcfb8100f67a705294d54f62f1.webp "Take B")
- [ ] Take A: the wordmark is correct and its model gave the largest native resolution
- [x] Take B: its paint and its raking sunrise light match the reference and the brief
- [ ] Neither stands out: both wordmarks pass, so pick the faster model
> Once the wordmark passes, the decision moves to paint and light. Take A's paint turned cooler and silvery under flat light, which matters for a car sold on its color.

? A colleague ran the shoot-out but rewrote the prompt for each model to get its best result. What do you do with the scorecard?
- [x] Reject the comparison and rerun with one prompt, word for word, for every model
- [ ] Accept it: each model was shown at its best, which is fairer
- [ ] Accept it, as long as each model also ran at its highest quality setting
> Prompts tuned per model are in Reject the comparison when. A fair shoot-out changes one thing, the model.
:::

## Make it repeatable

- **Keep a test pack per job** ([Keep sources as data](https://learn.scenario.com/browse)). The prompt, the reference images, the settings and the scorecard template in one collection. Rerunning the shoot-out when a model ships is then a 10 minute task. See [Keep up with new models](https://learn.scenario.com/courses/pick-the-right-model).
- **Build the shoot-out as a workflow.** In the **Node Editor**: an **Image** input (`Reference`), a **Text** input (`Test prompt`), and four **Image Generator** nodes, one per candidate, all fed by the same two inputs. Duplicate it per job; change only the models.
- **Write the decision where people work.** Put "Product scenes: GPT Image 2.5 Sunburst; drafts: Flare" in the **Project Instructions** so Scenario Agents follow it, and set it as the model in the team's App. See [Instructions and memory](https://learn.scenario.com/agents/instructions-and-memory.md).
- **Restrict when needed.** If the decision must be enforced, admins can limit the models available to the organization. See [Control model access](https://learn.scenario.com/courses/enterprise-admin).
- **Date every decision.** Models change monthly; a scorecard without a date is an opinion.

## Runs and time

| Step | Model | Runs or takes | Kept | Wait |
|---|---|---|---|---|
| Candidate 1 | GPT Image 2.5 Sunburst, high | 1 | 1 (winner) | about 40 s |
| Candidate 2 | GPT Image 2.5 Flare, high | 1 | 1 (runner-up) | about 27 s |
| Candidate 3 | Nano Banana 2.1, 2K | 1 | 0 | about 35 s |
| Candidate 4 | Seedream 5.0 Pro, 2048x1360 | 1 | 0 | about 126 s |
| Second seed, top two | Sunburst and Flare, high | 2 | 2 | about 40 s and 27 s |
| Grid, crops, scorecard | local | 1 | 1 | about 10 min of review |

Started in parallel, the four candidates finish in the time of the slowest one. Each candidate costs one run; read each price on the **Generate** button, or from a dry-run quote through the Scenario MCP, and copy it into the scorecard.

## Variations

- **Video:** same method with a fixed start frame and motion prompt; score drift, motion direction and artifacts frame by frame.
- **3D:** fixed input image; score topology, texture and silhouette in a viewer, plus export compatibility.
- **Audio:** fixed script or music brief; score with a transcript, a listen on headphones and the duration.
- **Upscalers:** fixed low-res input; score text, faces and invented detail at 1:1.
- **Cost-first teams:** add a "cost per accepted image" row: cost per run divided by the share of runs you would keep.

## If it breaks

- **Every model looks good.** The test is too easy. Use the hardest real brief: small text, hands, a product with a precise color or fine detail.
- **The results flip on a second run.** The models are close on this job. Pick on cost or speed, and note both as acceptable.
- **A model needs different settings to work.** Give each its own best quality setting, but never a different prompt; record the settings in the scorecard.
- **The team ignores the decision.** Put the winner in the App, the workflow and the project instructions, not only in a document.
- **A new model appears the week after.** Rerun the test pack with the new model against the current winner only.

Source: https://learn.scenario.com/playbooks/compare-models-for-your-studio