Captioned Speech gives duplicate text

[manascb1344]

Describe the bug
The captioned speech example gives duplicate word level timestamps

python kokoro-docker.py
Generating captioned speech for example texts...


Example 1:
Input text: Hello world! Welcome to the captioned speech system.
Response status: 200
Response headers: {'date': 'Sat, 08 Feb 2025 16:40:28 GMT', 'server': 'uvicorn', 'content-disposition': 'attachment; filename=speech.wav', 'x-accel-buffering': 'no', 'cache-control': 'no-cache', 'x-word-timestamps': '[{"word": "Hello", "start_time": 0.175, "end_time": 0.525}, {"word": "Hello", "start_time": 0.175, "end_time": 0.525}, {"word": "world", "start_time": 0.525, "end_time": 0.9}, {"word": "world", "start_time": 0.525, "end_time": 0.9}, {"word": "!", "start_time": 0.9, "end_time": 0.9875}, {"word": "!", "start_time": 0.9, "end_time": 0.9875}, {"word": "Welcome", "start_time": 0.9875, "end_time": 1.45}, {"word": "Welcome", "start_time": 0.9875, "end_time": 1.45}, {"word": "to", "start_time": 1.45, "end_time": 1.5375}, {"word": "to", "start_time": 1.45, "end_time": 1.5375}, {"word": "the", "start_time": 1.5375, "end_time": 1.625}, {"word": "the", "start_time": 1.5375, "end_time": 1.625}, {"word": "captioned", "start_time": 1.625, "end_time": 2.075}, {"word": "captioned", "start_time": 1.625, "end_time": 2.075}, {"word": "speech", "start_time": 2.075, "end_time": 2.4}, {"word": "speech", "start_time": 2.075, "end_time": 2.4}, {"word": "system", "start_time": 2.4, "end_time": 3.1}, {"word": "system", "start_time": 2.4, "end_time": 3.1}, {"word": ".", "start_time": 3.1, "end_time": 3.25}, {"word": ".", "start_time": 3.1, "end_time": 3.25}]', 'content-type': 'audio/wav', 'Transfer-Encoding': 'chunked'}
Audio saved to: [REMOVED]
Timestamps saved to:  [REMOVED]

Word-level timestamps:
Hello: 0.175s - 0.525s
Hello: 0.175s - 0.525s
world: 0.525s - 0.900s
world: 0.525s - 0.900s
!: 0.900s - 0.988s
!: 0.900s - 0.988s
Welcome: 0.988s - 1.450s
Welcome: 0.988s - 1.450s
to: 1.450s - 1.538s
to: 1.450s - 1.538s
the: 1.538s - 1.625s
the: 1.538s - 1.625s
captioned: 1.625s - 2.075s
captioned: 1.625s - 2.075s
speech: 2.075s - 2.400s
speech: 2.075s - 2.400s
system: 2.400s - 3.100s
system: 2.400s - 3.100s
.: 3.100s - 3.250s
.: 3.100s - 3.250s

Example 2:
Input text: The quick brown fox jumps over the lazy dog.
Response status: 200
Response headers: {'date': 'Sat, 08 Feb 2025 16:40:28 GMT', 'server': 'uvicorn', 'content-disposition': 'attachment; filename=speech.wav', 'x-accel-buffering': 'no', 'cache-control': 'no-cache', 'x-word-timestamps': '[{"word": "The", "start_time": 0.175, "end_time": 0.25}, {"word": "The", "start_time": 0.175, "end_time": 0.25}, {"word": "quick", "start_time": 0.25, "end_time": 0.5}, {"word": "quick", "start_time": 0.25, "end_time": 0.5}, {"word": "brown", "start_time": 0.5, "end_time": 0.8375}, {"word": "brown", "start_time": 0.5, "end_time": 0.8375}, {"word": "fox", "start_time": 0.8375, "end_time": 1.2375}, {"word": "fox", "start_time": 0.8375, "end_time": 1.2375}, {"word": "jumps", "start_time": 1.2375, "end_time": 1.5375}, {"word": "jumps", "start_time": 1.2375, "end_time": 1.5375}, {"word": "over", "start_time": 1.5375, "end_time": 1.7375}, {"word": "over", "start_time": 1.5375, "end_time": 1.7375}, {"word": "the", "start_time": 1.7375, "end_time": 1.825}, {"word": "the", "start_time": 1.7375, "end_time": 1.825}, {"word": "lazy", "start_time": 1.825, "end_time": 2.2}, {"word": "lazy", "start_time": 1.825, "end_time": 2.2}, {"word": "dog", "start_time": 2.2, "end_time": 2.85}, {"word": "dog", "start_time": 2.2, "end_time": 2.85}, {"word": ".", "start_time": 2.85, "end_time": 3.025}, {"word": ".", "start_time": 2.85, "end_time": 3.025}]', 'content-type': 'audio/wav', 'Transfer-Encoding': 'chunked'}
Audio saved to:  [REMOVED]
Timestamps saved to: [REMOVED]

Word-level timestamps:
The: 0.175s - 0.250s
The: 0.175s - 0.250s
quick: 0.250s - 0.500s
quick: 0.250s - 0.500s
brown: 0.500s - 0.838s
brown: 0.500s - 0.838s
fox: 0.838s - 1.238s
fox: 0.838s - 1.238s
jumps: 1.238s - 1.538s
jumps: 1.238s - 1.538s
over: 1.538s - 1.738s
over: 1.538s - 1.738s
the: 1.738s - 1.825s
the: 1.738s - 1.825s
lazy: 1.825s - 2.200s
lazy: 1.825s - 2.200s
dog: 2.200s - 2.850s
dog: 2.200s - 2.850s
.: 2.850s - 3.025s
.: 2.850s - 3.025s

Screenshots or console output

Branch / Deployment used
Using Docker Run command for the cpu

docker run -p 8880:8880 ghcr.io/remsky/kokoro-fastapi-cpu:v0.2.0post3

Operating System
On Linux (Pop OS) Ubuntu 22.04

[remsky]

Fix should be up shortly. Switched the endpoint to more closely mirror the openai endpoint streaming, with a tempfile link at the end for the captions.

[RBEmerson970]

I'm feeling mildly brain dead on finding later versions of v0.2.0postnn - I guessed my way to post4. Where do I get the right version name to at least past into docker run --gpus all -p 8880:8880 ghcr.io/remsky/kokoro-fastapi-gpu:v0.2.0post4

…

On Sun, Feb 9, 2025 at 10:57 PM remsky ***@***.***> wrote: Fix should be up shortly. Switched the endpoint to more closely mirror the openai endpoint streaming, with a tempfile link at the end for the captions. — Reply to this email directly, view it on GitHub <#139 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BFI36ZEFMEBX6JZRH6O4PXD2PAPR7AVCNFSM6AAAAABWX3LDFOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMNBWHA3DENZYGE> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[remsky]

Sorry, hadn't built just yet which is why the tag wasn't there, though it was on master, but I'll send it through as a release on v0.2.1 now, has a few other updates as well. Can track the build status here if it isn't up yet https://github.com/remsky/Kokoro-FastAPI/actions/runs/13234254893/job/36936577912

Can find the list of currently build images here. I do still carry a latest tag mostly for my own or dev use.

https://github.com/remsky/Kokoro-FastAPI/packages

[remsky]

https://github.com/remsky/Kokoro-FastAPI/releases/tag/v0.2.1 Let me know how it works for you