Skip to content
@tiny-aya-simultaneous-translation

tiny-aya-simultaneous-translation

Turkish ⇄ Hindi speech-to-speech translation: adapting a text-pretrained Cohere2 backbone to drive a frozen Moshi depth decoder over Mimi codes.

TinyAya — Turkish ⇄ Hindi Speech-to-Speech Translation

Can a text-pretrained multilingual model be taught to translate speech to speech, for a language pair with almost no parallel speech data?

We built the whole pipeline to find out: generate a synthetic parallel-speech corpus, encode it to neural audio tokens, and adapt a Cohere2 backbone — driving a frozen Moshi depth decoder over Mimi codes — with a text "inner-monologue" running alongside the audio stream.

What we found

The full-corpus run covered 76,250 steps / 2.07 epochs on a Cloud TPU v6e-16. Capability arrived in a clear order — language identity → text translation → audio synthesis — and the run bought the first two:

measured @best, 500 held-out rows, greedy hi→tr tr→hi
free-run text chrF++ — the inner-monologue 25.7 25.1
generated-audio ASR-chrF++ 3.7 9.6
ground-truth-audio ceiling, same pipeline 92.1 86.6

The model learns the translation mapping; intelligible audio synthesis is the frontier. That ceiling was measured through the same codec and ASR, so the gap is not a broken harness — it is the frozen depth decoder. We publish every score next to its ceiling, because a low number without one is unreadable.

📊 Full evaluation report · ✍️ Write-up · 📈 Training run · 🤗 Models & datasets

Repositories

repo what it is
model Start here. Stage-2 training, the 8-stage evaluation harness, and the TPU launch tooling. Dual-backend: the same code runs on GPU or TPU.
codec-finetuning Fine-tunes Mimi / DualCodec / Kanade on Turkish and Hindi across 8 optimizers, with Bayesian sweeps and bootstrap error bars — asking whether a language-specific codec beats the stock multilingual one.
sound-quality-check 4-stage speech-dataset quality control: round-trip ASR, MOS, SNR, VAD. This is what filtered the synthetic corpus.
hindi-tts-probe 5 Hindi TTS × 3 ASR round-trip benchmark. It picked the Hindi ASR judge used in the v0.3 evaluation.
tinyaya-moshi-backbone The proof-of-concept that validated Tiny Aya as a Moshi backbone — prior art for everything above.
dhruv Exploratory work and contributions.

Some repos in this organization are private (data-pipeline, tiny-aya-deliverables, simultaneous-translation) and are intentionally not linked — the corpora they produced are published on 🤗 Hugging Face instead.

From text to model

tr-hi-parallel-text          text triples (en pivot -> tr / hi)
        |  TTS
tr-hi-parallel-speech-v2     synthetic speech + round-trip ASR quality signals
        |  Mimi encode
tr-hi-mimi-encoded           8 codebooks @ 12.5 Hz + word-level alignments
        |  Stage-2 training
tr-hi-s2st-v0.3              the released model

Every dataset in that chain is public, as is the released model with its ~89-checkpoint suite — so the whole trajectory is inspectable, not just the endpoint.

Honest limitations

Training speech is entirely synthetic TTS, and the model is measurably distribution-bound: on real human speech the text score collapses from ~25 to ~8 chrF++. Translation references are machine-translated and carry their own ceiling. These are research artifacts for low-resource speech translation, not production translators. Earlier versions are kept with their mistakes disclosed — including two releases that trained audio-only because a filename check wrongly concluded the corpus had no text alignments. It ships 840,426.

Licensing

Code is Apache-2.0. Released model weights are CC-BY-NC-4.0 — non-commercial, inherited from CohereLabs/tiny-aya-base. Corpus text derives from FLORES (CC BY-SA 4.0 — share-alike propagates), OPUS-100 and TTS output. Check the individual dataset cards before redistributing.

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

Popular repositories Loading

  1. model model Public

    TinyAya Stage 2: TR↔HI speech-to-speech translation at scale

    Python 2 1

  2. tinyaya-moshi-backbone tinyaya-moshi-backbone Public

    Python 1 1

  3. dhruv dhruv Public

    Jupyter Notebook 1

  4. codec-finetuning codec-finetuning Public

    Python 1 2

  5. sound-quality-check sound-quality-check Public

    4-stage speech dataset quality control pipeline for ASR fine-tuning

    Python 1

  6. hindi-tts-probe hindi-tts-probe Public

    Hindi TTS quality probe: 5 TTS x 3 ASR benchmark on lahaja-eval

    Python

Repositories

Showing 7 of 7 repositories

People

This organization has no public members. You must be a member to see who’s a part of this organization.

Top languages

Loading…

Most used topics

Loading…