Model reference · open weights
whisper-tiny.en.tflite is an open-weight audio or speech model from nyadla-sys. AxForge deploys and operates it for you on dedicated EU-owned hardware — with the licence handled where one is required.
Available as managed deployment — configured and operated for you on dedicated EU hardware, quoted per deployment.
What it is
| Released by | nyadla-sys |
|---|---|
| Type | Audio & music |
| Task | Speech→text |
| Runs with | transformers |
| Released | 2023-09-03 |
| Popularity | 1k downloads / month |
| Licence | Open weights |
About
Converts OpenAI Whisper speech recognition models to TFLite format for on-device inference (e.g. Android), and generates the mel filter + vocab binary file needed by native C++ runtimes.
Dependencies are installed automatically on first run.
The script supports three modes depending on the --language argument:
.en models)python3.9 whisper_tflite_model_generation_and_test.py --model whisper-tiny.en
whisper-tiny.en.tfliteserving_default (transcribe English)forced_decoder_ids: [[2, 50359], [3, 50363]]python3.9 whisper_tflite_model_generation_and_test.py --model whisper-base --language fr
python3.9 whisper_tflite_model_generation_and_test.py --model whisper-base --language de
whisper-base.fr.tflite, whisper-base.de.tfliteserving_default (transcribe the specified language)forced_decoder_ids: [[1, ], [2, 50359], [3, 50363]]python3.9 whisper_tflite_model_generation_and_test.py --model whisper-base --language auto
whisper-base-transcribe-translate.tfliteserving_default (= transcribe), serving_transcribe, serving_translateforced_decoder_ids (transcribe): [[2, 50359], [3, 50363]]forced_decoder_ids (translate): [[2, 50358], [3, 50363]]# Default: whisper-tiny.en (English-only)
python3.9 whisper_tflite_model_generation_and_test.py
# Single-language French model
python3.9 whisper_tflite_model_generation_and_test.py --model whisper-base --language fr
# Transcribe-translate model (auto-detect language)
python3.9 whisper_tflite_model_generation_and_test.py --model whisper-base --language auto
# Test translate signature specifically
python3.9 whisper_tflite_model_generation_and_test.py --model whisper-small --language auto --task translate
# Show all options
python3.9 whisper_tflite_model_generation_and_test.py --help
| Argument | Default | Description |
|---|---|---|
--model | whisper-tiny.en | Whisper model to convert |
--language | en | Language code (en, fr, de, ...) or auto for auto-detection |
--task | transcribe | transcribe or translate. Only relevant with --language auto |
--multilingual / --no-multilingual | --multilingual | Vocab binary type. Auto-set to --no-multilingual for .en models |
| Model | Type | Parameters | Required VRAM | Relative Speed |
|---|---|---|---|---|
whisper-tiny.en | English-only | ~39M | ~1 GB | ~10x |
whisper-tiny | Multilingual | ~39M | ~1 GB | ~10x |
whisper-base.en | English-only | ~74M | ~1 GB | ~7x |
whisper-base | Multilingual | ~74M | ~1 GB | ~7x |
whisper-small.en | English-only | ~244M | ~2 GB | ~4x |
whisper-small | Multilingual | ~244M | ~2 GB | ~4x |
whisper-medium.en | English-only | ~769M | ~5 GB | ~2x |
whisper-medium | Multilingual | ~769M | ~5 GB | ~2x |
whisper-large | Multilingual | ~1550M | ~10 GB | 1x |
whisper-large-v3 | Multilingual | ~1550M | ~10 GB | 1x |
whisper-turbo | Multilingual | ~809M | ~6 GB | ~8x |
en, fr, hi, ko, de, zh, ja, es, ar, ru, pt, it, nl, sv, pl, da, fi, and many more.
Use auto for language auto-detection. See the full list.
| Mode | Signatures | Description |
|---|---|---|
English-only (.en) | serving_default | Transcribe English |
| Single-language | serving_default | Transcribe the forced language |
Transcribe-translate (auto) | serving_default, serving_transcribe, serving_translate | Auto-detect language, transcribe or translate to English |
Token reference:
50358 = ``50359 = ``50363 = ``| Step | Description |
|---|---|
| 0 | Install/verify Python dependencies |
| 1 | Configure model parameters and fetch decoder token mappings |
| 2 | Load the Whisper model and run a test transcription (English: LibriSpeech, other languages: Google FLEURS) |
| 3 | Patch TFForceTokensLogitsProcessor to avoid NaN values during TFLite export |
| 4 | Wrap the model with serving signature(s) and save as TF SavedModel |
| 5 | Convert the SavedModel to TFLite with dynamic range quantization |
| 6 | Verify the TFLite model produces correct output via the TFLite Interpreter |
| 7 | (Optional) Test the TFLite model against .wav audio files |
| 8 | Generate the mel filters + vocab binary file |
# English-only
whisper-tiny.en.tflite
filters_vocab_en.bin
# Single-language (e.g. French)
whisper-base.fr.tflite
filters_vocab_multilingual.bin
# Transcribe-translate (auto)
whisper-base-transcribe-translate.tflite
filters_vocab_multilingual.bin
These files are what you need for on-device Whisper inference on Android or other embedded platforms.
If you are using the Whisper tflite model, please cite:
@misc{nyadla-sys,
author={Niranjan Yadla},
title={{Whisper TFLite: OpenAI Whisper Model Port for Edge Devices}},
year=2022,
howpublished={GitHub Repository},
url={https://github.com/nyadla-sys/whisper.tflite},
url={https://github.com/moonshine-ai/openai-whisper},
note={Original TFLite implementation of OpenAI Whisper for on-device automatic speech recognition}
}
From the published model card. Full card on the HuggingFace links in the sidebar.
How it works
Using it via the API
Once AxForge deploys whisper-tiny-en-tflite for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (whisper-tiny-en-tflite below is illustrative; you get the exact model name on deployment.)
$ curl -sS https://api.axforge.ai/v1/audio/transcriptions \ -H "Authorization: Bearer $AXFORGE_API_KEY" \ -F model="whisper-tiny-en-tflite" -F file=@audio.mp3
Create an account — your API key is available in the console. 3M free tokens every 30 days with every new account.