---
language:
  - tr
  - hi
license: cc-by-sa-4.0
task_categories:
  - audio-to-audio
tags:
  - speech-to-speech-translation
  - mimi
  - moshi
  - audio-codec
  - turkish
  - hindi
pretty_name: "TR↔HI Mimi-Encoded Parallel Speech"
size_categories:
  - 1M<n<10M
---

# TR↔HI Mimi-Encoded Parallel Speech

Pre-encoded parallel Turkish↔Hindi speech pairs for training speech-to-speech translation models. All audio has been tokenized through the [Mimi](https://huggingface.co/kyutai/mimi) neural audio codec (8 codebooks, 12.5 Hz, 24kHz) and stored as `.pt` files with word-level text alignments.

## Dataset Summary

| | |
|---|---|
| **Source audio** | ~911 hours of synthetic parallel TR↔HI speech from [tr-hi-parallel-speech-v2](https://huggingface.co/datasets/tiny-aya-translate/tr-hi-parallel-speech-v2) |
| **TTS model** | [OmniVoice](https://github.com/k2-fsa/OmniVoice) (voice design mode, 14 voice designs) |
| **Text sources** | FLORES (~2K), OPUS-100 (~9K), machine-translated conversational datasets (~45K) |
| **Codec** | [Mimi](https://huggingface.co/kyutai/mimi) — 8 codebooks at 12.5 Hz |
| **Total encoded pairs** | ~1.24M train / ~65K val (full split) |
| **Quality-filtered subset** | ~25K train / ~500 val |
| **Directions** | Bidirectional: Turkish→Hindi and Hindi→Turkish |
| **QC** | Round-trip ASR validation with faster-whisper (large-v3), 86% pass rate at WER ≤ 0.20 |

## Files

### Encoded Audio (packed as tars)

Each tar contains `.pt` files and their corresponding alignment JSONs:

```
mimi_encoded_batch1_tars0-12.tar.gz
mimi_encoded_batch3.tar.gz
...
mimi_encoded_batch10.tar.gz
```

Each `.pt` file contains:

```python
{
    "src_codes": tensor([8, T_src]),   # source audio Mimi codes (8 codebooks × T frames)
    "tgt_codes": tensor([8, T_tgt]),   # target audio Mimi codes
    "src_text": str,                    # source language text
    "tgt_text": str,                    # target language text
    "src_lang": str, "tgt_lang": str,   # e.g. "tr" / "hi"
    "src_duration_s": float, "tgt_duration_s": float,
    "pair_id": str, "voice": str,
    "direction": str,                   # "tr->hi" or "hi->tr"
}
```

Each `.pt` file has two companion alignment JSONs, which map words to Mimi's
12.5 Hz frame grid for text-audio interleaving during training:

- `{stem}.src.alignments.json` — source-language word-level timestamps
- `{stem}.tgt.alignments.json` — target-language word-level timestamps

> ### ⚠️ Read this before writing a loader
>
> **The alignment paths inside `splits/*.jsonl` do not resolve as written.** The
> manifests were generated with an earlier naming scheme and record
> `encoded/{stem}_src.json`. The files this dataset actually ships are
> **`{stem}.src.alignments.json`** / **`{stem}.tgt.alignments.json`**, and they
> extract to the **data root** — *not* into `encoded/`.
>
> That is a double mismatch, in both the filename and the directory. It is not
> hypothetical: a checker that looked for the manifest's names found zero hits
> and the corpus was wrongly believed to have no text alignments at all, so two
> model releases trained audio-only before it was caught. There are **840,426 of
> each kind, at 100% coverage**.
>
> Resolve alignments by trying, in order: the manifest path as-is → `encoded/`
> + basename → `data_root/{stem}.{src,tgt}.alignments.json`. A reference
> implementation is `StreamingTranslationDataset._resolve_alignment` in the
> [model repo](https://github.com/tiny-aya-simultaneous-translation/model/blob/main/src/data/dataset.py).

### Splits

```
splits/
├── train.jsonl          # Full training split (~1.24M rows)
├── val.jsonl            # Full validation split (~65K rows)
├── train_26k.jsonl      # Quality-filtered 26K subset
└── val_500.jsonl        # Quality-filtered 500-sample val set
```

Each JSONL row:

```json
{
    "pt_path": "encoded/conv_13222_female_young_hi_rev.pt",
    "direction": "hi->tr",
    "pair_id": "conv_13222",
    "voice": "female_young_hi",
    "src_align_path": "encoded/conv_13222_female_young_hi_rev_src.json",
    "tgt_align_path": "encoded/conv_13222_female_young_hi_rev_tgt.json"
}
```

The two `*_align_path` values above are reproduced verbatim from the shipped
manifests **and are stale** — see the warning above for the paths that actually
exist on disk.

### 26K Subset Construction

The `train_26k.jsonl` and `val_500.jsonl` splits are a quality-filtered subset designed for initial training experiments under compute constraints. Construction criteria:

- **QC pass**: only pairs that passed round-trip ASR validation (WER ≤ 0.20)
- **Duration**: source + target between 2–30 seconds
- **Completeness**: both source and target `.pt` files and alignment JSONs must exist
- **Leak-free splitting**: rows are grouped by `pair_id` before splitting, so the same text pair in different voices always lands in the same split
- **Direction balance**: approximately equal TR→HI and HI→TR samples

The full 1.24M split contains all encoded pairs including those that didn't pass strict QC, for teams that want to apply their own filtering.

## Quality Control

Generated audio was validated using a round-trip ASR pipeline:

1. Transcribe each generated audio clip with [faster-whisper](https://github.com/SYSTRAN/faster-whisper) (large-v3)
2. Compute WER against the original text prompt
3. Accept if WER ≤ 0.20 and duration is between 0.5s–30s

Additional metrics computed (available in the [QC pipeline](https://github.com/tiny-aya-simultaneous-translation/sound-quality-check)):
- **DNSMOS** — perceptual audio quality (neural MOS predictor)
- **SNR** — signal-to-noise ratio
- **VAD speech ratio** — fraction of clip containing actual speech
- **CER** — character error rate

Overall pass rate: **86%** across validated shards. Failures concentrated in extreme voice designs (e.g., `male_old_deep`) and very long sentences.

## Voice Designs

14 OmniVoice voice designs used for speaker diversity:

```
default, female_young_hi, male_mid_hi, female_high, male_elderly,
female_mid, male_young, female_brit, male_deep, female_teen,
male_teen, male_old_deep, female_old_low, male_old_russian
```

## Usage

```python
from huggingface_hub import hf_hub_download
import json, torch

# Download a split
split = hf_hub_download(
    "tiny-aya-translate/tr-hi-mimi-encoded",
    "splits/train_26k.jsonl",
    repo_type="dataset"
)

# Load a sample
with open(split) as f:
    row = json.loads(f.readline())

# Download and load the .pt file
pt_file = hf_hub_download(
    "tiny-aya-translate/tr-hi-mimi-encoded",
    row["pt_path"],
    repo_type="dataset"
)
data = torch.load(pt_file, weights_only=False)
print(data["src_codes"].shape)  # [8, T_src]
print(data["tgt_codes"].shape)  # [8, T_tgt]
```

For training with the [model repo](https://github.com/tiny-aya-simultaneous-translation/model):

```bash
python scripts/train_hierarchical.py \
    --config configs/gpu/stage2_26k_parallel.yaml \
    --train_split splits/train_26k.jsonl \
    --val_split splits/val_500.jsonl \
    --encoded_dir encoded/
```

## Related

- [Model training code](https://github.com/tiny-aya-simultaneous-translation/model)
- Data generation pipeline — private repo, not publicly browsable
- [Audio QC pipeline](https://github.com/tiny-aya-simultaneous-translation/sound-quality-check)
- [Raw parallel speech (911h)](https://huggingface.co/datasets/tiny-aya-translate/tr-hi-parallel-speech-v2)
- [Text pairs](https://huggingface.co/datasets/tiny-aya-translate/tr-hi-parallel-text)

## Citation

```bibtex
@misc{tinyaya_tr_hi_mimi_encoded,
    title  = {TR-HI Mimi-Encoded Parallel Speech Dataset},
    author = {tiny-aya-translate},
    year   = {2026},
    url    = {https://huggingface.co/datasets/tiny-aya-translate/tr-hi-mimi-encoded}
}
```

## Where this sits

The v0.3 speech-to-speech pipeline, end to end:

```
tr-hi-parallel-text          text triples (en pivot -> tr / hi)
        |  TTS
tr-hi-parallel-speech-v2     synthetic speech + QC signals
        |  Mimi encode
tr-hi-mimi-encoded           8-codebook tokens + word alignments
        |  Stage-2 training
tr-hi-s2st-v0.3              the released model
```

| | |
|---|---|
| **Model** | [`tr-hi-s2st-v0.3`](https://huggingface.co/tiny-aya-translate/tr-hi-s2st-v0.3) |
| **Text** | [`tr-hi-parallel-text`](https://huggingface.co/datasets/tiny-aya-translate/tr-hi-parallel-text) |
| **Speech** | [`tr-hi-parallel-speech-v2`](https://huggingface.co/datasets/tiny-aya-translate/tr-hi-parallel-speech-v2) · [`-v3`](https://huggingface.co/datasets/tiny-aya-translate/tr-hi-parallel-speech-v3) |
| **Encoded** | [`tr-hi-mimi-encoded`](https://huggingface.co/datasets/tiny-aya-translate/tr-hi-mimi-encoded) |
| **Eval sets** | [`fleurs-tr-hi-mimi-encoded`](https://huggingface.co/datasets/tiny-aya-translate/fleurs-tr-hi-mimi-encoded) · [`lahaja-eval`](https://huggingface.co/datasets/tiny-aya-translate/lahaja-eval) · [`cv-tr-eval`](https://huggingface.co/datasets/tiny-aya-translate/cv-tr-eval) |

## Code

| repo | what it does |
|---|---|
| [`model`](https://github.com/tiny-aya-simultaneous-translation/model) | Stage-2 training, evaluation harness and TPU launch tooling |
| [`sound-quality-check`](https://github.com/tiny-aya-simultaneous-translation/sound-quality-check) | 4-stage speech-dataset quality control |

## Project

**TinyAya Stage 2** — Turkish⇄Hindi speech-to-speech translation with a text
inner-monologue: a LoRA-adapted Cohere2 backbone driving a **frozen** Moshi depth
decoder over Mimi codes.

The v0.3 run covered **76,250 steps / 2.07 epochs** on a Cloud TPU v6e-16
(best val composite **2.8199** @ step 76,000). Read honestly: the text
inner-monologue **learns to translate** (free-run chrF++ ~25.7 / 25.1), while
**intelligible audio synthesis remains the frontier** (ASR-chrF++ 3.7 / 9.6
against a 92.1 / 86.6 ground-truth-audio ceiling) — bounded by the frozen depth
decoder, not by translation understanding.

- **Results:** [v0.3 evaluation report](https://github.com/tiny-aya-simultaneous-translation/model/blob/main/docs/v0.3-eval-report.md)
- **Training run:** [W&B `xzcb60bl`](https://wandb.ai/cataluna84/tinyaya-stage2-tpu/runs/xzcb60bl) · [emergence report](https://wandb.ai/cataluna84/tinyaya-stage2-tpu/reports/TinyAya-v0.3-Emergence-and-Data-Efficiency--VmlldzoxNzU1OTU1NQ==)
- **Blog:** [Adapting Moshi for Low-Resource Speech Translation](https://labscommunity.cohere.com/blog/2026/adapting-moshi-low-resource-speech-translation/)

Compute for the v0.3 run was provided by **Google's TPU Research Cloud (TRC)**.