diff --git a/Cargo.lock b/Cargo.lock
index 6a3767f..e22f4a2 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -228,8 +228,7 @@ dependencies = [
 [[package]]
 name = "async-openai"
 version = "0.28.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "14d76e2f5af19477d6254415acc95ba97c6cc6f3b1e3cb4676b7f0fab8194298"
+source = "git+https://github.com/etkecc/async-openai?branch=async-openai-v0.28.1-patched#856953c2d4485342df625fd0525363362075e8a8"
 dependencies = [
  "async-openai-macros",
  "backoff",
@@ -254,8 +253,7 @@ dependencies = [
 [[package]]
 name = "async-openai-macros"
 version = "0.1.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0289cba6d5143bfe8251d57b4a8cac036adf158525a76533a7082ba65ec76398"
+source = "git+https://github.com/etkecc/async-openai?branch=async-openai-v0.28.1-patched#856953c2d4485342df625fd0525363362075e8a8"
 dependencies = [
  "proc-macro2",
  "quote",
diff --git a/Cargo.toml b/Cargo.toml
index d41bd37..4e75e62 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -17,7 +17,7 @@ path = "src/lib.rs"
 [dependencies]
 anthropic = "=0.0.8"
 anyhow = "1.0.*"
-async-openai = "0.28.*"
+async-openai = { git = "https://github.com/etkecc/async-openai", branch = "async-openai-v0.28.1-patched" }
 base64 = "0.22.*"
 chrono = { version = "0.4.*", default-features = false, features = ["std", "now"] }
 # We'd rather not depend on this, but we cannot use the ruma-events EventContent macro without it.
diff --git a/src/agent/provider/entity/image.rs b/src/agent/provider/entity/image.rs
index f85e367..b524f41 100644
--- a/src/agent/provider/entity/image.rs
+++ b/src/agent/provider/entity/image.rs
@@ -58,6 +58,7 @@ impl From<ImageSource> for async_openai::types::ImageInput {
         async_openai::types::ImageInput::from_vec_u8(
             value.filename,
             value.bytes,
+            value.mime_type.to_string(),
         )
     }
 }
diff --git a/src/agent/provider/openai/controller.rs b/src/agent/provider/openai/controller.rs
index 2dd19e9..e2e1299 100644
--- a/src/agent/provider/openai/controller.rs
+++ b/src/agent/provider/openai/controller.rs
@@ -6,7 +6,7 @@ use async_openai::{
     types::{
         ChatCompletionRequestMessage, CreateChatCompletionRequestArgs, CreateImageRequestArgs,
         CreateSpeechRequestArgs, CreateTranscriptionRequestArgs, CreateImageEditRequestArgs,
-        ImageInput, ImageModel, DallE2ImageSize, ImageResponseFormat, Image,
+        ImageModel, DallE2ImageSize, ImageResponseFormat, Image,
     },
 };
 
@@ -209,6 +209,7 @@ impl ControllerTrait for Controller {
             .file(async_openai::types::AudioInput::from_vec_u8(
                 filename,
                 media,
+                mime_type.to_string(),
             ))
             .language(language.clone())
             .build()?;
@@ -368,11 +369,14 @@ impl ControllerTrait for Controller {
             ));
         };
 
-        let Some(first_image) = images.into_iter().next() else {
+        if images.is_empty() {
             return Err(anyhow::anyhow!("No image sources provided"));
-        };
+        }
 
-        let image_input: ImageInput = first_image.into();
+        let mut image_inputs = Vec::new();
+        for image in images {
+            image_inputs.push(image.into());
+        }
 
         let dalle2_size = match image_generation_config.size {
             async_openai::types::ImageSize::S256x256 => Some(DallE2ImageSize::S256x256),
@@ -397,7 +401,7 @@ impl ControllerTrait for Controller {
         let mut request_builder = CreateImageEditRequestArgs::default();
 
         request_builder
-            .image(image_input)
+            .image(image_inputs)
             .prompt(prompt.to_owned())
             .model(model);