SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2408.16725 · 2024

Mini-Omni: Language Models Can Hear, Talk While Thinking in Streaming

arXiv · PDF · Open in the Atlas

Code that ran

11 of 14 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; 2 honoured the contract we drafted, 2 violated it and 0 have no contract verdict; 7 failures of our instrument; 3 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/2408.16725.json · badge: /api/ran/2408.16725.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
gpt-omni/mini-omni canonical (classifier label)
classifier basis not recorded on the link
11 ran · 3 no recorded run
FunctionStatusWhere it lives
apply_rope Ran
ran, violated the contract we drafted
gpt-omni/mini-omni/litgpt/model.py
code served (permissive licence) · get_code("aa32c6faaa562b69")
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 Tensor, ndim 3, dtype family float, shape (2, 3, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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

torch.manual_seed(0)

# Construct tensors for apply_rope function
# x: input tensor with shape (batch, seq_len, head_size)
# The function expects x to have a last dimension (head_size) that can be split in half
batch_size = 2
seq_len = 3
head_size = 4

x = torch.randn(batch_size, seq_len, head_size)

# cos and sin tensors should have compatible shape for broadcasting
# They should be able to broadcast with x
cos = torch.randn(batch_size, seq_len, head_size)
sin = torch.randn(batch_size, seq_len, head_size)

ARGS = {
    "x": x,
    "cos": cos,
    "sin": sin,
}
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 Tensor, ndim 3, dtype family float, shape (2, 3, 4) — shape (2, 3, 4) as a later run recorded it
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
build_mask_cache Ran
ran, violated the contract we drafted
gpt-omni/mini-omni/litgpt/model.py
code served (permissive licence) · get_code("7ee1c3ba71156110")
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 Tensor, ndim 4, dtype family float, shape (1, 1, 8, 8) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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

torch.manual_seed(0)

ARGS = {
    "max_seq_length": 8,
    "device": None,
}
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 Tensor, ndim 4, dtype family float, shape (1, 1, 8, 8) — shape (1, 1, 8, 8) as a later run recorded it
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
collect_chunks Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
gpt-omni/mini-omni/utils/vad.py
code served (permissive licence) · get_code("847a58dff80851e5")
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 ndarray, ndim 1, dtype family float, shape (6,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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 numpy as np

# Create a small audio array
# The function indexes this with chunk['start']:chunk['end']
audio = np.array([0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8], dtype=np.float32)

# Create a list of chunks that specify start and end indices
chunks = [
    {'start': 0, 'end': 2},
    {'start': 3, 'end': 5},
    {'start': 6, 'end': 8}
]

ARGS = {
    'audio': audio,
    'chunks': chunks
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family float, shape (6,) — shape (6,) as a later run recorded it
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
find_multiple Ran
ran, honoured the contract we drafted
gpt-omni/mini-omni/litgpt/utils.py
pointer only (licence: MIT); the stored copy get_code serves, from hrtan/fastmix, is pointer only (licence: NONE) · get_code("f5d9c92914e0ccfa")
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 hrtan/fastmix, 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.

find_resume_path Ran
ran; our contract draft was wrong, not the code
gpt-omni/mini-omni/litgpt/utils.py
code served (permissive licence) · get_code("9b5c939537b0e38c")
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 PosixPath. 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 tempfile
from pathlib import Path

# Create a temporary directory structure to simulate out_dir with checkpoint files
temp_dir = tempfile.TemporaryDirectory()
out_dir = Path(temp_dir.name)

# Create some checkpoint directories and files
step_1_dir = out_dir / "step-100"
step_1_dir.mkdir(parents=True, exist_ok=True)
(step_1_dir / "model.pth").touch()

step_2_dir = out_dir / "step-200"
step_2_dir.mkdir(parents=True, exist_ok=True)
(step_2_dir / "model.pth").touch()

# Test cases for different resume parameter values
ARGS = {
    "resume": "auto",
    "out_dir": out_dir,
}
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 PosixPath — a PosixPath output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a PosixPath 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
get_snac Ran
ran; our contract draft was wrong, not the code
gpt-omni/mini-omni/utils/snac_utils.py
code served (permissive licence) · get_code("641d60a367fa3083")
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 torch
import numpy as np

torch.manual_seed(0)

# Analyzing the function:
# - list_output: accessed as list_output[j][start - nums_generate - 5 + j + i]
#   This is a list of 7 sequences (j ranges 0-6)
#   Each sequence needs to be indexable
# - index: used as 'start = index'
# - nums_generate: used in range(nums_generate) and in index calculation

# The index calculation: start - nums_generate - 5 + j + i
# where start = index, j ranges 0-6, i ranges 0 to nums_generate-1
# We need to ensure all indices are valid

# Let's set:
# - nums_generate = 3
# - index = 15
# Then start = 15
# Min index accessed: 15 - 3 - 5 + 0 + 0 = 7
# Max index accessed: 15 - 3 - 5 + 6 + 2 = 15

# So we need sequences of length at least 16
# Create 7 sequences, each with 20 elements

list_output = [
    list(range(20 * i, 20 * i + 20)) for i in range(7)
]

index = 15
nums_generate = 3

ARGS = {
    "list_output": list_output,
    "index": index,
    "nums_generate": nums_generate,
}
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 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
init_out_dir Ran
ran; our contract draft was wrong, not the code
gpt-omni/mini-omni/litgpt/utils.py
code served (permissive licence) · get_code("615c2d6183fce709")
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 PosixPath. 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 os
from pathlib import Path

# Set up a temporary directory path for testing
out_dir = Path("/tmp/test_output")

ARGS = {
    "out_dir": out_dir
}
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 PosixPath — a PosixPath output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a PosixPath 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
layershift Ran
ran, honoured the contract we drafted
gpt-omni/mini-omni/utils/snac_utils.py
code served (permissive licence) · get_code("eba9e8fbcfaae4be")
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 int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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

torch.manual_seed(0)

# The function layershift takes:
# - input_id: some identifier/value
# - layer: layer index
# - stride: stride value (default 4160)
# - shift: shift value (default 152000)
# The function simply performs arithmetic: input_id + shift + layer * stride

# Since the function only uses these parameters for arithmetic operations,
# they should be numeric values (int or float)

ARGS = {
    "input_id": 1000,
    "layer": 5,
    "stride": 4160,
    "shift": 152000,
}
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 int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
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
multinomial_num_samples_1 Ran
ran; our contract draft was wrong, not the code
gpt-omni/mini-omni/litgpt/generate/base.py
code served (permissive licence) · get_code("17e4446536a73126")
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 Tensor, ndim 2, dtype family int, shape (2, 1) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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

torch.manual_seed(0)

# Create a valid probability tensor
# Shape: (batch_size=2, num_classes=4)
# Values should be non-negative and sum to 1 along the last dimension
probs = torch.tensor([
    [0.1, 0.2, 0.3, 0.4],
    [0.25, 0.25, 0.25, 0.25]
], dtype=torch.float32)

ARGS = {
    "probs": probs
}
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 Tensor, ndim 2, dtype family int, shape (2, 1) — shape (2, 1) as a later run recorded it
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
sample Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
gpt-omni/mini-omni/litgpt/generate/base.py
code served (permissive licence) · get_code("62b68fa657c0d6eb")
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 Tensor, ndim 1, dtype family int, shape (1,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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 torch.nn.functional as F

torch.manual_seed(0)

# Create a small logits tensor
# Shape: (batch_size=2, sequence_length=3, vocab_size=4)
# The function accesses logits[0, -1], so it expects at least 2D
# We make it 3D to match typical language model output
logits = torch.randn(2, 3, 4)

# Define parameters for the sample function
temperature = 0.8
top_k = 2
top_p = 0.9

ARGS = {
    "logits": logits,
    "temperature": temperature,
    "top_k": top_k,
    "top_p": top_p,
}
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 Tensor, ndim 1, dtype family int, shape (1,) — shape (1,) as a later run recorded it
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
sample_top_p Ran
ran; our contract draft was wrong, not the code
gpt-omni/mini-omni/litgpt/generate/base.py
code served (permissive licence) · get_code("a2d9cc3c4221f547")
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 Tensor, ndim 1, dtype family float, shape (4,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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

torch.manual_seed(0)

# Create a small logits tensor
# Shape: (4,) - a single sample with 4 vocabulary items
logits = torch.randn(4)

# Define top_p parameter (nucleus sampling threshold)
top_p = 0.9

ARGS = {
    "logits": logits,
    "top_p": top_p,
}
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 Tensor, ndim 1, dtype family float, shape (4,) — shape (4,) as a later run recorded it
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
build_rope_cache Not yet run
unverified (harvested, no recorded run)
gpt-omni/mini-omni/litgpt/model.py
code served (permissive licence) · get_code("c0cabd60c81402a5")
How it ran

no invocation recorded.

generate_audio_data Not yet run
unverified (harvested, no recorded run)
gpt-omni/mini-omni/utils/snac_utils.py
code served (permissive licence) · get_code("1d7eb4c029d371bd")
How it ran

no invocation recorded.

get_speech_timestamps Not yet run
unverified (harvested, no recorded run)
gpt-omni/mini-omni/utils/vad.py
code served (permissive licence) · get_code("f8e3df53ca30326f")
How it ran

no invocation recorded.

Also present in other papers' code

Of 14 samples lifted from this paper's linked repositories, 14 carry a function identity and 0 do not. Of the 14 distinct functions with an identity, 14 are also present in other papers whose code we hold (5 with at least one paper dated earlier, 3 the same month, 14 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
build_rope_cache 6 other papers
3 dated earlier, 3 later
2405.20340 (2024-05)
MotionLLM: Understanding Human Behaviors from Human Motions and Videos
2406.06046 (2024-06)
MATES : Model-Aware Data Selection for Efficient Pretraining with Data Influence Models
2406.10209 (2024-06)
Be like a Goldfish, Don't Memorize! Mitigating Memorization in Generative LLMs
and 3 more not named here
find_multiple 6 other papers
2 dated earlier, 1 the same month, 3 later
2307.09288 (2023-07)
Llama 2: Open Foundation and Fine-Tuned Chat Models
2406.06046 (2024-06)
MATES : Model-Aware Data Selection for Efficient Pretraining with Data Influence Models
2408.15793 (2024-08)
Language Adaptation on a Tight Academic Compute Budget: Tokenizer Swapping Works and Pure bfloat16 Is Enough
and 3 more not named here
apply_rope 5 other papers
2 dated earlier, 3 later
2406.06046 (2024-06)
MATES : Model-Aware Data Selection for Efficient Pretraining with Data Influence Models
2406.10209 (2024-06)
Be like a Goldfish, Don't Memorize! Mitigating Memorization in Generative LLMs
2410.08102 (2024-10)
Efficient Pretraining Data Selection for Language Models via Multi-Actor Collaboration
and 2 more not named here
build_mask_cache 5 other papers
2 dated earlier, 3 later
2406.06046 (2024-06)
MATES : Model-Aware Data Selection for Efficient Pretraining with Data Influence Models
2406.10209 (2024-06)
Be like a Goldfish, Don't Memorize! Mitigating Memorization in Generative LLMs
2410.08102 (2024-10)
Efficient Pretraining Data Selection for Language Models via Multi-Actor Collaboration
and 2 more not named here
multinomial_num_samples_1 3 other papers
1 dated earlier, 2 later
2406.06046 (2024-06)
MATES : Model-Aware Data Selection for Efficient Pretraining with Data Influence Models
2410.08102 (2024-10)
Efficient Pretraining Data Selection for Language Models via Multi-Actor Collaboration
2410.11190 (2024-10)
Mini-Omni2: Towards Open-source GPT-4o with Vision, Speech and Duplex Capabilities
collect_chunks 2 other papers
1 the same month, 1 later
2408.01800 (2024-08)
MiniCPM-V: A GPT-4V Level MLLM on Your Phone
2410.11190 (2024-10)
Mini-Omni2: Towards Open-source GPT-4o with Vision, Speech and Duplex Capabilities
get_speech_timestamps 2 other papers
1 the same month, 1 later
2408.01800 (2024-08)
MiniCPM-V: A GPT-4V Level MLLM on Your Phone
2410.11190 (2024-10)
Mini-Omni2: Towards Open-source GPT-4o with Vision, Speech and Duplex Capabilities
find_resume_path 1 other paper
1 later
2410.11190 (2024-10)
Mini-Omni2: Towards Open-source GPT-4o with Vision, Speech and Duplex Capabilities
generate_audio_data 1 other paper
1 later
2410.11190 (2024-10)
Mini-Omni2: Towards Open-source GPT-4o with Vision, Speech and Duplex Capabilities
get_snac 1 other paper
1 later
2410.11190 (2024-10)
Mini-Omni2: Towards Open-source GPT-4o with Vision, Speech and Duplex Capabilities
init_out_dir 1 other paper
1 later
2410.11190 (2024-10)
Mini-Omni2: Towards Open-source GPT-4o with Vision, Speech and Duplex Capabilities
layershift 1 other paper
1 later
2410.11190 (2024-10)
Mini-Omni2: Towards Open-source GPT-4o with Vision, Speech and Duplex Capabilities
sample 1 other paper
1 later
2410.11190 (2024-10)
Mini-Omni2: Towards Open-source GPT-4o with Vision, Speech and Duplex Capabilities
sample_top_p 1 other paper
1 later
2410.11190 (2024-10)
Mini-Omni2: Towards Open-source GPT-4o with Vision, Speech and Duplex Capabilities

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

Recent advances in language models have achieved significant progress. GPT-4o, as a new milestone, has enabled real-time conversations with humans, demonstrating near-human natural fluency. Such human-computer interaction necessitates models with the capability to perform reasoning directly with the audio modality and generate output in streaming. However, this remains beyond the reach of current academic models, as they typically depend on extra TTS systems for speech synthesis, resulting in undesirable latency. This paper introduces the Mini-Omni, an audio-based end-to-end conversational model, capable of real-time speech interaction. To achieve this capability, we propose a text-instructed speech generation method, along with batch-parallel strategies during inference to further boost the performance. Our method also helps to retain the original model's language capabilities with minimal degradation, enabling other works to establish real-time interaction capabilities. We call this training method "Any Model Can Talk". We also introduce the VoiceAssistant-400K dataset to fine-tune models optimized for speech output. To our best knowledge, Mini-Omni is the first fully end-to-end, open-source model for real-time speech interaction, offering valuable potential for future research.

For agents

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

get_harvested_code_for_paper("2408.16725")
get_code_for_paper("2408.16725")
have("2408.16725")

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

curl https://syntology.ai/api/ran/2408.16725.json

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

[![Syntology run record](https://syntology.ai/api/ran/2408.16725.svg)](https://syntology.ai/paper/2408.16725)

Connect an agent — have() is free.