feat(diffusers): add support for wan2.2

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

core/backend/video.go

@@ -7,7 +7,7 @@ import (
 	model "github.com/mudler/LocalAI/pkg/model"
 )
 
-func VideoGeneration(height, width int32, prompt, startImage, endImage, dst string, loader *model.ModelLoader, modelConfig config.ModelConfig, appConfig *config.ApplicationConfig) (func() error, error) {
+func VideoGeneration(height, width int32, prompt, negativePrompt, startImage, endImage, dst string, numFrames, fps, seed int32, cfgScale float32, step int32, loader *model.ModelLoader, modelConfig config.ModelConfig, appConfig *config.ApplicationConfig) (func() error, error) {
 
 	opts := ModelOptions(modelConfig, appConfig)
 	inferenceModel, err := loader.Load(
@@ -22,12 +22,18 @@ func VideoGeneration(height, width int32, prompt, startImage, endImage, dst stri
 		_, err := inferenceModel.GenerateVideo(
 			appConfig.Context,
 			&proto.GenerateVideoRequest{
-				Height:     height,
-				Width:      width,
-				Prompt:     prompt,
-				StartImage: startImage,
-				EndImage:   endImage,
-				Dst:        dst,
+				Height:         height,
+				Width:          width,
+				Prompt:         prompt,
+				NegativePrompt: negativePrompt,
+				StartImage:     startImage,
+				EndImage:       endImage,
+				NumFrames:      numFrames,
+				Fps:            fps,
+				Seed:           seed,
+				CfgScale:       cfgScale,
+				Step:           step,
+				Dst:            dst,
 			})
 		return err
 	}

core/http/endpoints/localai/video.go

@@ -61,7 +61,7 @@ func downloadFile(url string) (string, error) {
 */
 // VideoEndpoint
 // @Summary Creates a video given a prompt.
-// @Param request body schema.OpenAIRequest true "query params"
+// @Param request body schema.VideoRequest true "query params"
 // @Success 200 {object} schema.OpenAIResponse "Response"
 // @Router /video [post]
 func VideoEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, appConfig *config.ApplicationConfig) func(c *fiber.Ctx) error {
@@ -166,7 +166,23 @@ func VideoEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, appConfi
 
 		baseURL := c.BaseURL()
 
-		fn, err := backend.VideoGeneration(height, width, input.Prompt, src, input.EndImage, output, ml, *config, appConfig)
+		fn, err := backend.VideoGeneration(
+			height,
+			width,
+			input.Prompt,
+			input.NegativePrompt,
+			src,
+			input.EndImage,
+			output,
+			input.NumFrames,
+			input.FPS,
+			input.Seed,
+			input.CFGScale,
+			input.Step,
+			ml,
+			*config,
+			appConfig,
+		)
 		if err != nil {
 			return err
 		}

core/schema/localai.go

@@ -28,6 +28,7 @@ type GalleryResponse struct {
 type VideoRequest struct {
 	BasicModelRequest
 	Prompt         string  `json:"prompt" yaml:"prompt"`
+	NegativePrompt string  `json:"negative_prompt" yaml:"negative_prompt"`
 	StartImage     string  `json:"start_image" yaml:"start_image"`
 	EndImage       string  `json:"end_image" yaml:"end_image"`
 	Width          int32   `json:"width" yaml:"width"`
@@ -36,6 +37,7 @@ type VideoRequest struct {
 	FPS            int32   `json:"fps" yaml:"fps"`
 	Seed           int32   `json:"seed" yaml:"seed"`
 	CFGScale       float32 `json:"cfg_scale" yaml:"cfg_scale"`
+	Step           int32   `json:"step" yaml:"step"`
 	ResponseFormat string  `json:"response_format" yaml:"response_format"`
 }