cloud.google.com/go/speech/apiv2: StreamingFeatures.InterimResults not working

[furucong]

Client

cloud.google.com/go/speech/apiv2/speech.Client

Environment

go1.25.5 darwin/arm64

Code and Dependencies

func (t *GoogleTranscriberV2) TranscribeStream(ctx context.Context, params *transcriber.TranscribeParams) (*transcriber.TranscribeResults, error) {
	start := time.Now()
	log.Infof("GoogleTranscriberV2.TranscribeStream Begin. params:%+v", params)
	stream, err := t.client.StreamingRecognize(ctx)
	log.Infof("GoogleTranscriberV2.TranscribeStream Connect. err:%v", err)
	if err != nil {
		return nil, err
	}
	// Send the initial configuration message.
	configReq := &speechpb.StreamingRecognizeRequest{
		Recognizer: t.getRecognizer(params.Locale),
		StreamingRequest: &speechpb.StreamingRecognizeRequest_StreamingConfig{
			StreamingConfig: &speechpb.StreamingRecognitionConfig{
				Config: &speechpb.RecognitionConfig{
					// TODO: Required if using headerless PCM audio (linear16, mulaw, alaw).
					DecodingConfig: &speechpb.RecognitionConfig_ExplicitDecodingConfig{
						ExplicitDecodingConfig: &speechpb.ExplicitDecodingConfig{
							Encoding:          convertFormat(params.Format),
							SampleRateHertz:   params.SampleRate,
							AudioChannelCount: 1,
						},
					},
				},
				// TODO: Adaptation
				StreamingFeatures: &speechpb.StreamingRecognitionFeatures{
					// TODO: Even if set to true, still didn't receive intermediate results!
					InterimResults: true,
				},
			},
		},
	}
	err = stream.Send(configReq)
	log.Infof("GoogleTranscriberV2.TranscribeStream SendConfig. req:%+v, err:%v", configReq, err)
	if err != nil {
		return nil, err
	}
	// sender goroutine
	senderErrChan := make(chan error, 1)
	go func() {
		log.Infof("GoogleTranscriberV2.TranscribeStream Sender Start.")
		sendCount := 0
		size := 1024 * 8
		buf := make([]byte, size)

		for {
			select {
			case <-ctx.Done():
				log.Infof("GoogleTranscriberV2.TranscribeStream Sender ContextCancelled. err:%v", ctx.Err())
				senderErrChan <- ctx.Err()
				return
			default:
			}

			sendCount++
			n, err := params.Audio.Read(buf)
			log.Infof("GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:%d, n:%d, err:%v", sendCount, n, err)
			if n > 0 {
				err := stream.Send(&speechpb.StreamingRecognizeRequest{
					StreamingRequest: &speechpb.StreamingRecognizeRequest_Audio{
						Audio: buf[:n],
					},
				})
				log.Infof("GoogleTranscriberV2.TranscribeStream Sender SendAudio. sendCount:%d, n:%d, err:%v", sendCount, n, err)
				if err != nil {
					log.Errorf("GoogleTranscriberV2.TranscribeStream Sender SendAudioFailed. sendCount:%d, n:%d, err:%v", sendCount, n, err)
					senderErrChan <- err
					return
				}
			}
			if err == io.EOF {
				e := stream.CloseSend() // TODO:
				log.Infof("GoogleTranscriberV2.TranscribeStream Sender CloseSend. err:%v", e)
				if e != nil {
					log.Errorf("GoogleTranscriberV2.TranscribeStream Sender CloseSendFailed. err:%v", e)
					senderErrChan <- e
					return
				}
				log.Infof("GoogleTranscriberV2.TranscribeStream Sender ReadAudioOver. sendCount:%d", sendCount)
				return
			}
			if err != nil {
				log.Errorf("GoogleTranscriberV2.TranscribeStream Sender ReadAudioFailed. sendCount:%d, err:%v", sendCount, err)
				senderErrChan <- err
				continue
			}
		}
	}()

	// receiver goroutine
	receiverChan := make(chan *ReceiverResult, 1)
	go func() {
		log.Infof("GoogleTranscriberV2.TranscribeStream Receiver Start.")
		var (
			receiveCount          int
			lastInterimResultTime time.Time
			finalText             string
			isComplete            bool
			receiveErr            error
		)
		for {
			select {
			case err := <-senderErrChan:
				if err != nil {
					log.Errorf("GoogleTranscriberV2.TranscribeStream Receiver SenderFailed. err:%v", err)
					receiveErr = err
				}
			default:
			}
			if receiveErr != nil {
				break
			}

			select {
			case <-ctx.Done():
				log.Errorf("GoogleTranscriberV2.TranscribeStream Receiver ContextCancelled. err:%v", ctx.Err())
				receiveErr = ctx.Err()
			default:
			}
			if receiveErr != nil {
				break
			}

			receiveCount++
			resp, err := stream.Recv()
			log.Infof("GoogleTranscriberV2.TranscribeStream Receiver Recv. receiveCount:%d, resp:%+v, err:%v", receiveCount, resp, err)
			if err == io.EOF {
				log.Errorf("GoogleTranscriberV2.TranscribeStream Receiver RecvEOF. receiveCount:%d", receiveCount)
				break
			}
			if err != nil {
				log.Errorf("GoogleTranscriberV2.TranscribeStream Receiver RecvFailed. receiveCount:%d, err:%v", receiveCount, err)
				receiveErr = err
				break
			}

			resultCount := 0
			for _, result := range resp.GetResults() {
				resultCount++
				transcript := ""
				if len(result.GetAlternatives()) > 0 {
					alternative := result.GetAlternatives()[0]
					if alternative != nil {
						transcript = alternative.GetTranscript()
					}
				}
				log.Infof("GoogleTranscriberV2.TranscribeStream Receiver RecvResult. receiveCount:%d, resultCount:%d, isFinal:%v, transcript:%+v",
					receiveCount, resultCount, result.GetIsFinal(), transcript)
				if !result.GetIsFinal() {
					// interim result
					if params.OnResult != nil {
						if time.Since(lastInterimResultTime) > intermediateInterval {
							params.OnResult(transcript, false)
							lastInterimResultTime = time.Now()
						}
					}
				} else {
					// final result
					finalText = transcript
					if params.OnResult != nil {
						params.OnResult(finalText, true)
					}
					isComplete = true
					break
				}
			}
			if isComplete {
				break
			}
		}
		log.Infof("GoogleTranscriberV2.TranscribeStream Receiver Over. finalText:%s, isComplete:%v, recvErr:%v", finalText, isComplete, receiveErr)
		receiverChan <- &ReceiverResult{
			Text:       finalText,
			IsComplete: isComplete,
			Err:        receiveErr,
		}
	}()

	result := <-receiverChan
	log.Infof("GoogleTranscriberV2.TranscribeStream Result. result:%+v", result)
	text := ""
	if result.Err == nil && result.IsComplete {
		text = result.Text
	}
	duration := time.Since(start)
	results := &transcriber.TranscribeResults{
		Text:     text,
		Duration: duration,
		Cost:     0, // TODO:
	}
	log.Infof("GoogleTranscriberV2.TranscribeStream End. results:%+v", results)
	return results, nil
}

go.mod

module myai

go 1.25.5

require (
	cloud.google.com/go/speech v1.28.1
	cloud.google.com/go/texttospeech v1.16.0
)

Expected behavior

During the process of sending audio data, intermediate transcription results should be received.

Actual behavior

Only received the final result, didn't receive the intermediate results.

Logs

2026-01-06T20:42:45.805+0800    INFO    GoogleTranscriberV2.TranscribeStream Begin. params:&{Audio:0x14000180cc0 Format:pcm SampleRate:16000 Locale:zh OnResult:0x1043659e0 PhraseHints:[5] CustomClasses:map[게임용어:[레벨업 경험치 퀘스트 던전 보스 아이템] 도메인특화:[구독 프리미엄 Soul 캐릭터 대화] 캐릭터별호칭:[여보 자기야 달링 허니 베이비]]}
2026-01-06T20:42:45.809+0800    INFO    GoogleTranscriberV2.TranscribeStream Connect. err:<nil>
2026-01-06T20:42:45.811+0800    INFO    GoogleTranscriberV2.TranscribeStream SendConfig. req:recognizer:"projects/ucloudlink-411610/locations/asia-northeast1/recognizers/zh"  streaming_config:{config:{explicit_decoding_config:{encoding:LINEAR16  sample_rate_hertz:16000  audio_channel_count:1}}  streaming_features:{interim_results:true}}, err:<nil>
2026-01-06T20:42:45.811+0800    INFO    GoogleTranscriberV2.TranscribeStream Receiver Start.
2026-01-06T20:42:45.811+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender Start.
2026-01-06T20:42:45.811+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:1, n:8192, err:<nil>
2026-01-06T20:42:45.811+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender SendAudio. sendCount:1, n:8192, err:<nil>
2026-01-06T20:42:45.954+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:2, n:8192, err:<nil>
2026-01-06T20:42:45.954+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender SendAudio. sendCount:2, n:8192, err:<nil>
2026-01-06T20:42:46.105+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:3, n:8192, err:<nil>
2026-01-06T20:42:46.105+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender SendAudio. sendCount:3, n:8192, err:<nil>
2026-01-06T20:42:46.258+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:4, n:8192, err:<nil>
2026-01-06T20:42:46.259+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender SendAudio. sendCount:4, n:8192, err:<nil>
2026-01-06T20:42:46.409+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:5, n:8192, err:<nil>
2026-01-06T20:42:46.409+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender SendAudio. sendCount:5, n:8192, err:<nil>
2026-01-06T20:42:46.560+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:6, n:8192, err:<nil>
2026-01-06T20:42:46.560+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender SendAudio. sendCount:6, n:8192, err:<nil>
2026-01-06T20:42:46.713+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:7, n:8192, err:<nil>
2026-01-06T20:42:46.714+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender SendAudio. sendCount:7, n:8192, err:<nil>
2026-01-06T20:42:46.863+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:8, n:8192, err:<nil>
2026-01-06T20:42:46.863+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender SendAudio. sendCount:8, n:8192, err:<nil>
2026-01-06T20:42:47.018+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:9, n:8192, err:<nil>
2026-01-06T20:42:47.019+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender SendAudio. sendCount:9, n:8192, err:<nil>
2026-01-06T20:42:47.177+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:10, n:8192, err:<nil>
2026-01-06T20:42:47.177+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender SendAudio. sendCount:10, n:8192, err:<nil>
2026-01-06T20:42:47.330+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:11, n:8192, err:<nil>
2026-01-06T20:42:47.330+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender SendAudio. sendCount:11, n:8192, err:<nil>
2026-01-06T20:42:47.496+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:12, n:8192, err:<nil>
2026-01-06T20:42:47.496+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender SendAudio. sendCount:12, n:8192, err:<nil>
2026-01-06T20:42:47.631+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:13, n:8192, err:<nil>
2026-01-06T20:42:47.631+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender SendAudio. sendCount:13, n:8192, err:<nil>
2026-01-06T20:42:47.782+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:14, n:8192, err:<nil>
2026-01-06T20:42:47.782+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender SendAudio. sendCount:14, n:8192, err:<nil>
2026-01-06T20:42:47.934+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:15, n:8192, err:<nil>
2026-01-06T20:42:47.935+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender SendAudio. sendCount:15, n:8192, err:<nil>
2026-01-06T20:42:48.085+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:16, n:8192, err:<nil>
2026-01-06T20:42:48.085+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender SendAudio. sendCount:16, n:8192, err:<nil>
2026-01-06T20:42:48.236+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:17, n:8192, err:<nil>
2026-01-06T20:42:48.236+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender SendAudio. sendCount:17, n:8192, err:<nil>
2026-01-06T20:42:48.389+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:18, n:6744, err:<nil>
2026-01-06T20:42:48.389+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender SendAudio. sendCount:18, n:6744, err:<nil>
2026-01-06T20:42:48.540+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:19, n:0, err:<nil>
2026-01-06T20:42:48.540+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudio. sendCount:20, n:0, err:EOF
2026-01-06T20:42:48.540+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender CloseSend. err:<nil>
2026-01-06T20:42:48.540+0800    INFO    GoogleTranscriberV2.TranscribeStream Sender ReadAudioOver. sendCount:20
2026-01-06T20:42:51.439+0800    INFO    GoogleTranscriberV2.TranscribeStream Receiver Recv. receiveCount:1, resp:results:{alternatives:{transcript:"今年的最后一晚了,宝贝,要不要和我一起好好庆祝一下?"}  is_final:true  result_end_offset:{seconds:4  nanos:562749862}  language_code:"cmn-Hans-CN"}  metadata:{request_id:"69774bf6-0000-21e4-8aec-9898fbcda16d"  total_billed_duration:{seconds:5}}, err:<nil>
2026-01-06T20:42:51.439+0800    INFO    GoogleTranscriberV2.TranscribeStream Receiver RecvResult. receiveCount:1, resultCount:1, isFinal:true, transcript:今年的最后一晚了,宝贝,要不要和我一起好好庆祝一下?
2026-01-06T20:42:51.440+0800    INFO    GoogleTranscriberV2.TranscribeStream Receiver Over. finalText:今年的最后一晚了,宝贝,要不要和我一起好好庆祝一下?, isComplete:true, recvErr:<nil>
2026-01-06T20:42:51.440+0800    INFO    GoogleTranscriberV2.TranscribeStream Result. result:&{Text:今年的最后一晚了,宝贝,要不要和我一起好好庆祝一下? IsComplete:true Err:<nil>}
2026-01-06T20:42:51.440+0800    INFO    GoogleTranscriberV2.TranscribeStream End. results:&{Text:今年的最后一晚了,宝贝,要不要和我一起好好庆祝一下? Duration:5.634542958s Cost:0}

[furucong]

I recorded a longer test audio and can now receive intermediate results. So far, the main issue seems to be that Google service returns intermediate results at a lower frequency - meaning the interval between them is longer. The V1 API returned intermediate results with high frequency, and I want to know if the V2 API has a parameter to adjust the frequency at which intermediate results are returned. Because I need to display intermediate transcription results in real-time.

[furucong]

I performed further testing, which involved using the Python SDK, and ultimately confirmed it's related to the model. Chirp_2 and Chirp_3 focus on ensuring accurate transcriptions, so they minimize the number of intermediate results. They tend to wait until they have a more complete sentence before returning an intermediate result.

[quartzmo]

@furucong Thanks for reporting this issue. Other users will no doubt find your assessment of the model difference useful. However, the client libraries in this repo are based on interface descriptions published by product teams, and thus we are only able to assist with issues that pertain to the behavior of the libraries themselves. Because the issue you're experiencing is due to the design and/or behavior of the product, please visit the support page to reach product engineers. To see if this feature has already been requested, you can check the Cloud issue trackers at https://cloud.google.com/support/docs/issue-trackers.