Add support for aspect ratio in gemini image generation #3412

mwildehahn · 2025-11-13T16:41:12Z

Fix for #3119

DouweM · 2025-11-13T20:04:38Z

docs/builtin-tools.md


 _(This example is complete, it can be run "as is")_

+To control the aspect ratio when using Gemini image models, include the `ImageGenerationTool` explicitly:


Move this example under "Configuration Options" please

DouweM · 2025-11-13T20:05:36Z

docs/builtin-tools.md

 |----------|-----------|-------|
 | OpenAI Responses | ✅ | Full feature support. Only supported by models newer than `gpt-5`. Metadata about the generated image, like the [`revised_prompt`](https://platform.openai.com/docs/guides/tools-image-generation#revised-prompt) sent to the underlying image model, is available on the [`BuiltinToolReturnPart`][pydantic_ai.messages.BuiltinToolReturnPart] that's available via [`ModelResponse.builtin_tool_calls`][pydantic_ai.messages.ModelResponse.builtin_tool_calls]. |
-| Google | ✅ | No parameter support. Only supported by [image generation models](https://ai.google.dev/gemini-api/docs/image-generation) like `gemini-2.5-flash-image`. These models do not support [structured output](output.md) or [function tools](tools.md). These models will always generate images, even if this built-in tool is not explicitly specified. |
+| Google | ✅ | Supports the `aspect_ratio` parameter when explicitly provided. Only supported by [image generation models](https://ai.google.dev/gemini-api/docs/image-generation) like `gemini-2.5-flash-image`. These models do not support [structured output](output.md) or [function tools](tools.md) and will always generate images, even if this built-in tool is not explicitly specified. |


I'd say "Limited parameter support" like we do further up for web search.

pydantic_ai_slim/pydantic_ai/builtin_tools.py

DouweM · 2025-11-13T20:06:48Z

pydantic_ai_slim/pydantic_ai/builtin_tools.py

+
+    Supported by:
+
+    * Google image-generation models (Gemini) when the tool is explicitly enabled.


I think we can drop "when the tool is explicitly enabled." as that's implied by this being on that builtin tool class.

I wonder if we could support some of these values for OpenAI as well by mapping to one of the size options. Then we'd only need to raise an error from OpenAI if another value is used, or if size and aspect_ratio are used at the same time.

DouweM · 2025-11-13T20:07:33Z

pydantic_ai_slim/pydantic_ai/models/google.py

                        )
+                    if tool.aspect_ratio:
+                        if image_config and image_config.get('aspect_ratio') != tool.aspect_ratio:
+                            raise UserError(


We really only support a single instance anyway, so we can drop this and just always set image_config

DouweM · 2025-11-13T20:07:57Z

pydantic_ai_slim/pydantic_ai/models/google.py

-            response_schema=response_schema,
-            response_modalities=modalities,
-        )
+        config: GenerateContentConfigDict = {


Why did we have to change how this is built?

DouweM

@mwildehahn Thanks for doing the OpenAI Responses support, that's a nice quality of life detail. Just 2 points on docs and then we'll merge!

docs/builtin-tools.md

DouweM · 2025-11-14T14:50:55Z

docs/builtin-tools.md

 | `partial_images` | ✅ | ❌ |
 | `quality` | ✅ | ❌ |
 | `size` | ✅ | ❌ |
+| `aspect_ratio` | ✅ (1:1, 2:3, 3:2) | ✅ |


Suggested change

| `aspect_ratio` | ✅ (1:1, 2:3, 3:2) | ✅ |

| `aspect_ratio` | ✅ (`1:1`, `2:3`, `3:2`) | ✅ |

Ideally we'd use a different emoji for partial support; any ideas? 😄

github-actions · 2025-11-22T14:00:34Z

This PR is stale, and will be closed in 3 days if no reply is received.

DouweM · 2025-12-01T22:03:03Z

@mwildehahn Reminder to look at 2 small changes here!

svilupp · 2025-12-03T21:10:12Z

Excited for this feature! Thank you for putting it together

mwildehahn · 2025-12-09T17:11:41Z

@DouweM done

DouweM · 2025-12-10T00:01:46Z

pydantic_ai_slim/pydantic_ai/models/google.py

            tool_config=tool_config,
            response_mime_type=response_mime_type,
-            response_json_schema=response_schema,
+            response_schema=response_schema,


Please revert this change!

DouweM · 2025-12-10T00:05:13Z

Thanks @mwildehahn! You accidentally changed response_json_schema to response_schema thought which caused a few tests to fail 😅 @ajac-zero also fixed up your PR yesterday in #3672 which is essentially identical but does pass CI, so I'll merge that one. I'll make sure you're credited in the changelog :)

mwildehahn · 2025-12-10T01:29:16Z

Ah sorry! All good, sorry I was slow on this.

…

On Tue, Dec 9, 2025 at 4:05 PM Douwe Maan ***@***.***> wrote: *DouweM* left a comment (pydantic/pydantic-ai#3412) <#3412 (comment)> Thanks @mwildehahn <https://github.com/mwildehahn>! You accidentally changed response_json_schema to response_schema thought which caused a few tests to fail 😅 @ajac-zero <https://github.com/ajac-zero> also fixed up your PR yesterday in #3672 <#3672> which is essentially identical but does pass CI, so I'll merge that one. I'll make sure you're credited in the changelog :) — Reply to this email directly, view it on GitHub <#3412 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAFMAUMQBATOZZK4WJ2K2OL4A5PU7AVCNFSM6AAAAACMAW4BTGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTMMZUG44TEMJSGA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

mwildehahn mentioned this pull request Nov 13, 2025

Support for aspect ratio control in image generation (gemini's nano-banana) #3119

Closed

mwildehahn force-pushed the mh/aspect-ratio-control branch from e7b6dec to 9a3d8b0 Compare November 13, 2025 16:52

DouweM requested changes Nov 13, 2025

View reviewed changes

DouweM self-assigned this Nov 13, 2025

DouweM added the awaiting author revision label Nov 13, 2025

DouweM requested changes Nov 14, 2025

View reviewed changes

github-actions bot added the Stale label Nov 22, 2025

DouweM removed the Stale label Nov 24, 2025

DouweM mentioned this pull request Nov 27, 2025

Support gemini-3-pro-image-preview with Nano Banana Pro #3576

Merged

DouweM mentioned this pull request Dec 1, 2025

Add OpenRouter image generation support #3599

Open

ajac-zero mentioned this pull request Dec 8, 2025

Add support for aspect ratio in gemini image generation #3672

Merged

mwildehahn added 5 commits December 9, 2025 08:58

Add support for aspect ratio in gemini image generation

0f76b55

Update snapshot

779a864

Add test to examples

7f0d68e

Review feedback

8720f29

Review feedback re: openai

f2f2cd6

mwildehahn force-pushed the mh/aspect-ratio-control branch from d09bb8b to f2f2cd6 Compare December 9, 2025 17:11

DouweM reviewed Dec 10, 2025

View reviewed changes

DouweM closed this Dec 10, 2025


		_(This example is complete, it can be run "as is")_

		To control the aspect ratio when using Gemini image models, include the `ImageGenerationTool` explicitly:


		Supported by:

		* Google image-generation models (Gemini) when the tool is explicitly enabled.

	\| `aspect_ratio` \| ✅ (1:1, 2:3, 3:2) \| ✅ \|
	\| `aspect_ratio` \| ✅ (`1:1`, `2:3`, `3:2`) \| ✅ \|

Add support for aspect ratio in gemini image generation #3412

Add support for aspect ratio in gemini image generation #3412

Uh oh!

Conversation

mwildehahn commented Nov 13, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

DouweM left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Nov 22, 2025

Uh oh!

DouweM commented Dec 1, 2025

Uh oh!

svilupp commented Dec 3, 2025

Uh oh!

mwildehahn commented Dec 9, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

DouweM commented Dec 10, 2025

Uh oh!

mwildehahn commented Dec 10, 2025 via email

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants