【大模型文生图、文生音频实战Demo】基于Spring AI Alibaba和阿里百炼大模型实现文生图、文生视频

大模型实现文生图

使用阿里百炼大模型-wanx2.1-t2i-turbo，作为文生图调用的大模型

• 代码

@RestController
public class ImageModelController {
    private final ImageModel imageModel;

    ImageModelController(@Qualifier("dashScopeImageModel") ImageModel imageModel) {
        this.imageModel = imageModel;
    }

    @RequestMapping("/image")
    public String image(String input) {
        ImageOptions options = ImageOptionsBuilder.builder()
                .model("wanx2.1-t2i-turbo")
                .height(1024)
                .width(1024)
                .build();

        ImagePrompt imagePrompt = new ImagePrompt(input, options);
        ImageResponse response = imageModel.call(imagePrompt);
        String imageUrl = response.getResult().getOutput().getUrl();

        return "redirect:" + imageUrl;
    }
}

model指定模型，再指定生成图片的宽高

• postman调用接口
  
  点击链接打开，下载图片可查看
  

雨西湖，雷峰塔哈哈哈有品位

大模型实现文生音频

• 代码

@RestController
@RequestMapping("/audio")
public class AudioModelController {

    private final SpeechSynthesisModel speechSynthesisModel;

    @Autowired
    public AudioModelController(SpeechSynthesisModel speechSynthesisModel) {
        this.speechSynthesisModel = speechSynthesisModel;
    }

    @GetMapping("/synthesize")
    public ResponseEntity<byte[]> synthesizeSpeech(@RequestParam String text) throws IOException {
        // 构建语音合成请求
        SpeechSynthesisPrompt prompt = new SpeechSynthesisPrompt(text);

        // 调用模型生成语音
        SpeechSynthesisResponse response = speechSynthesisModel.call(prompt);
        ByteBuffer audioData = response.getResult().getOutput().getAudio();

        // 将 ByteBuffer 转换为字节数组
        byte[] audioBytes = new byte[audioData.remaining()];
        audioData.get(audioBytes);

        // 返回音频流（MP3格式）
        return ResponseEntity.ok()
                .contentType(MediaType.APPLICATION_OCTET_STREAM)
                .header("Content-Disposition", "attachment; filename=output.mp3")
                .body(audioBytes);
    }


}

• 网页访问调用

http://localhost:8080/audio/synthesize?text=你好

下载后可以得到对应文本的音频

以上，完成了用Spring AI Alibaba和调用各类大模型的方式，实现了文生图和文生音频！