SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2303.05309 · ICCV · 2023

MixSpeech: Cross-Modality Self-Learning with Audio-Visual Stream Mixup for Visual Speech Translation and Recognition

Zhou Zhao, Ye Wang, Rongjie Huang, Huadai Liu, Zehan Wang, Linjun Li, Aoxiong Yin, Tao Jin, Xize Cheng, Wang Lin

arXiv · PDF · Open in the Atlas

Code that ran

7 of 7 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 6 failures of our instrument; 0 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2303.05309.json · badge: /api/ran/2303.05309.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
rongjiehuang/transpeech community (classifier label)
classifier basis not recorded on the link
7 ran · 0 no recorded run
FunctionStatusWhere it lives
discriminator_loss Ran
ran; our contract draft was wrong, not the code
rongjiehuang/transpeech/research/TranSpeech/hifigan/models.py
pointer only (licence: MIT); the stored copy get_code serves, from zjzser/wmcodec, is pointer only (licence: NONE) · get_code("7137577cbef51217")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from zjzser/wmcodec, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

extract_audio_for_eval Ran
ran; our contract draft was wrong, not the code
rongjiehuang/transpeech/research/TranSpeech/asr_bleu/compute_asr_bleu.py
code served (permissive licence) · get_code("5961cc3df73025cb")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tempfile
import os
from pathlib import Path

# Create a temporary directory with mock audio files
tmpdir = tempfile.mkdtemp()

# Create some dummy audio files matching the expected naming pattern
for i in range(3):
    audio_file = Path(tmpdir) / f"{i}_pred.wav"
    audio_file.touch()

ARGS = {
    "audio_dirpath": tmpdir,
    "audio_format": "n_pred.wav",
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
feature_loss Ran
ran, honoured the contract we drafted
rongjiehuang/transpeech/research/TranSpeech/hifigan/models.py
pointer only (licence: MIT); the stored copy get_code serves, from zjzser/wmcodec, is pointer only (licence: NONE) · get_code("e453b51f0ed5fb28")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from zjzser/wmcodec, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

generator_loss Ran
ran; our contract draft was wrong, not the code
rongjiehuang/transpeech/research/TranSpeech/hifigan/models.py
pointer only (licence: MIT); the stored copy get_code serves, from zjzser/wmcodec, is pointer only (licence: NONE) · get_code("1a9d74439d969cfd")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from zjzser/wmcodec, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

merge_tailo_init_final Ran
ran; our contract draft was wrong, not the code
rongjiehuang/transpeech/research/TranSpeech/asr_bleu/compute_asr_bleu.py
code served (permissive licence) · get_code("dddc74f9be0a9b20")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

def merge_tailo_init_final(text):
    """
    Hokkien ASR hypothesis post-processing.
    """
    sps = text.strip().split()
    results = []
    last_syllable = ''
    for sp in sps:
        if sp == 'NULLINIT' or sp == 'nullinit':
            continue
        last_syllable += sp
        if sp[-1].isnumeric():
            results.append(last_syllable)
            last_syllable = ''
    if last_syllable != '':
        results.append(last_syllable)
    return ' '.join(results)


ARGS = {
    'text': 'ka1 NULLINIT ma2 nullinit bo5'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
remove_tone Ran
ran; our contract draft was wrong, not the code
rongjiehuang/transpeech/research/TranSpeech/asr_bleu/compute_asr_bleu.py
code served (permissive licence) · get_code("fce0a244cd01d4fd")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# The function remove_tone takes a single parameter: text
# It expects a string where words are space-separated and each word ends with a tone marker
# The function removes the last character (tone) from each word

text = "ma1 bo2 kiann3 kua7"

ARGS = {
    "text": text
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
retrieve_asr_config Ran
ran; our contract draft was wrong, not the code
rongjiehuang/transpeech/research/TranSpeech/asr_bleu/utils.py
code served (permissive licence) · get_code("d538507264c24d0b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import json
import tempfile
import os

# Create a temporary JSON file with ASR model configs
temp_dir = tempfile.mkdtemp()
json_path = os.path.join(temp_dir, "asr_config.json")

config_data = {
    "english": {
        "v1": {
            "model_name": "wav2vec2",
            "sample_rate": 16000,
            "num_mels": 80,
            "hidden_size": 768
        },
        "v2": {
            "model_name": "wav2vec2-large",
            "sample_rate": 16000,
            "num_mels": 80,
            "hidden_size": 1024
        }
    },
    "spanish": {
        "v1": {
            "model_name": "xlsr_53",
            "sample_rate": 16000,
            "num_mels": 80,
            "hidden_size": 768
        }
    }
}

with open(json_path, 'w') as f:
    json.dump(config_data, f)

ARGS = {
    "lang_key": "english",
    "asr_version": "v1",
    "json_path": json_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 distinct functions with an identity, 7 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 7 later) and 0 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
feature_loss 15 other papers
10 dated earlier, 5 later
2006.04558 (2020-06)
FastSpeech 2: Fast and High-Quality End-to-End Text to Speech
2106.06103 (2021-06)
Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech
2106.07889 (2021-06)
UnivNet: A Neural Vocoder with Multi-Resolution Spectrogram Discriminators for High-Fidelity Waveform Generation
and 12 more not named here
discriminator_loss 13 other papers
8 dated earlier, 5 later
2006.04558 (2020-06)
FastSpeech 2: Fast and High-Quality End-to-End Text to Speech
2106.07889 (2021-06)
UnivNet: A Neural Vocoder with Multi-Resolution Spectrogram Discriminators for High-Fidelity Waveform Generation
2109.13821 (2021-09)
Diffusion-Based Voice Conversion with Fast Maximum Likelihood Sampling Scheme
and 10 more not named here
generator_loss 13 other papers
9 dated earlier, 4 later
2006.04558 (2020-06)
FastSpeech 2: Fast and High-Quality End-to-End Text to Speech
2106.06103 (2021-06)
Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech
2106.07889 (2021-06)
UnivNet: A Neural Vocoder with Multi-Resolution Spectrogram Discriminators for High-Fidelity Waveform Generation
and 10 more not named here
extract_audio_for_eval 1 other paper
1 later
2406.03049 (2024-06)
StreamSpeech: Simultaneous Speech-to-Speech Translation with Multi-task Learning
merge_tailo_init_final 1 other paper
1 later
2406.03049 (2024-06)
StreamSpeech: Simultaneous Speech-to-Speech Translation with Multi-task Learning
remove_tone 1 other paper
1 later
2406.03049 (2024-06)
StreamSpeech: Simultaneous Speech-to-Speech Translation with Multi-task Learning
retrieve_asr_config 1 other paper
1 later
2406.03049 (2024-06)
StreamSpeech: Simultaneous Speech-to-Speech Translation with Multi-task Learning

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Multi-media communications facilitate global interaction among people. However, despite researchers exploring cross-lingual translation techniques such as machine translation and audio speech translation to overcome language barriers, there is still a shortage of cross-lingual studies on visual speech. This lack of research is mainly due to the absence of datasets containing visual speech and translated text pairs. In this paper, we present AVMuST-TED, the first dataset for Audio-Visual Multilingual Speech Translation, derived from TED talks. Nonetheless, visual speech is not as distinguishable as audio speech, making it difficult to develop a mapping from source speech phonemes to the target language text. To address this issue, we propose MixSpeech, a cross-modality self-learning framework that utilizes audio speech to regularize the training of visual speech tasks. To further minimize the cross-modality gap and its impact on knowledge transfer, we suggest adopting mixed speech, which is created by interpolating audio and visual streams, along with a curriculum learning strategy to adjust the mixing ratio as needed. MixSpeech enhances speech translation in noisy environments, improving BLEU scores for four languages on AVMuST-TED by +1.4 to +4.2. Moreover, it achieves state-of-the-art performance in lip reading on CMLR (11.1%), LRS2 (25.5%), and LRS3 (28.0%).

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2303.05309")
get_code_for_paper("2303.05309")
have("2303.05309")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2303.05309.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2303.05309.svg)](https://syntology.ai/paper/2303.05309)

Connect an agent — have() is free.