speech-nano/README.md

---
license: apache-2.0
language:
- en
tags:
- text-to-speech
- tts
- speech-synthesis
- pytorch
- ultra-small
- local-tts
- efficient-inference
- experimental
pipeline_tag: text-to-speech
library_name: pytorch
---

<p align="center">
  <img src="https://huggingface.co/owensong/Inflect-Nano-v1/resolve/main/assets/inflect-nano-banner.svg" alt="Inflect-Nano-v1 banner" width="100%">
</p>

# Inflect-Nano-v1

**Inflect-Nano-v1 is a 4.63M-parameter English text-to-speech model, including its neural vocoder.**

It is built around one question: how usable can a complete local TTS stack get when the entire inference model is smaller than many single embeddings tables in modern speech systems?

This is not a SOTA voice model. The point is the size-to-functionality ratio: a complete text-to-waveform TTS pipeline that is small enough for tiny local experiments, offline assistants, embedded prototypes, browser/WASM-style work, and efficient inference research.

<p align="center">
  <b>4.632M total params</b> | <b>3.465M acoustic</b> | <b>1.167M vocoder</b> | <b>24 kHz</b> | <b>English</b>
</p>

## Listen First

These examples cover punctuation, short questions, numbers, hard words, longer phrasing, and transition-heavy prompts.

| Prompt | Audio |
|---|---|
| "Did the timing change?" she answered. "Then why did Logan leave?" | <audio controls preload="none" src="https://huggingface.co/owensong/Inflect-Nano-v1/resolve/main/examples/example_01.wav"></audio> |
| Who puts a parking meter next to an ER label? | <audio controls preload="none" src="https://huggingface.co/owensong/Inflect-Nano-v1/resolve/main/examples/example_02.wav"></audio> |
| Please say neighborhood, statistics, and anesthesiologist clearly, without rushing through the middle syllables. | <audio controls preload="none" src="https://huggingface.co/owensong/Inflect-Nano-v1/resolve/main/examples/example_03.wav"></audio> |
| I said 91, not 306, which is a very different number. | <audio controls preload="none" src="https://huggingface.co/owensong/Inflect-Nano-v1/resolve/main/examples/example_04.wav"></audio> |
| The inference path looked natural, but the decoder still needed a smoother transition before Marcus approved the final test. | <audio controls preload="none" src="https://huggingface.co/owensong/Inflect-Nano-v1/resolve/main/examples/example_05.wav"></audio> |
| The appointment moved to 1:25, the invoice was $674.96, and the archive was labeled 1998. | <audio controls preload="none" src="https://huggingface.co/owensong/Inflect-Nano-v1/resolve/main/examples/example_06.wav"></audio> |
| If Logan sounded uneasy, then it happened near Long Beach, and the pause has to carry that. | <audio controls preload="none" src="https://huggingface.co/owensong/Inflect-Nano-v1/resolve/main/examples/example_07.wav"></audio> |
| The word aluminum should not steal attention from the softer ending after entrepreneur. | <audio controls preload="none" src="https://huggingface.co/owensong/Inflect-Nano-v1/resolve/main/examples/example_08.wav"></audio> |

## Why This Exists

Most strong TTS models are large. Many "small" TTS demos also depend on a larger external vocoder, which hides a lot of the real inference cost. Inflect-Nano-v1 counts both sides:

```text
text -> acoustic model -> mel spectrogram -> neural vocoder -> waveform
```

The model is intentionally constrained:

| Capability | Inflect-Nano-v1 |
|---|---:|
| Complete text-to-waveform stack | Yes |
| External vocoder required | No |
| Total inference parameters | **4.632M** |
| Sample rate | **24 kHz** |
| Language | English |
| Voice count | 1 |
| Voice cloning | No |
| Multilingual | No |

## Size Context

Approximate parameter comparison:

| Model / class | Params | Relative to Inflect-Nano-v1 |
|---|---:|---:|
| Inflect-Nano-v1 | **4.63M** | 1.0x |
| Kokoro-82M | ~82M | ~17.7x larger |
| Chatterbox-sized TTS | ~500M | ~108x larger |
| Multi-billion TTS systems | 1B+ | 216x+ larger |

The comparison is about model scale, not quality parity. Larger models are expected to sound better.

## Quickstart

```bash
git clone https://huggingface.co/owensong/Inflect-Nano-v1
cd Inflect-Nano-v1
pip install -r requirements.txt
```

Generate audio:

```bash
python inference.py \
  --text "Wait, are you actually being for real now? I can't believe it!" \
  --out sample.wav
```

CPU:

```bash
python inference.py \
  --device cpu \
  --text "Please say neighborhood, statistics, and anesthesiologist clearly." \
  --out sample_cpu.wav
```

Optional controls:

```bash
python inference.py \
  --text "No, seriously, did Jordan leave the receipt in Albuquerque?" \
  --length-scale 1.03 \
  --pitch-scale 1.00 \
  --energy-scale 1.00 \
  --out sample_controlled.wav
```

Local Gradio demo:

```bash
python app.py
```

The text frontend may download tokenizer files on first run.

## Model Summary

| Component | Parameters | Notes |
|---|---:|---|
| Acoustic model | **3,465,125** | Compact non-autoregressive FastSpeech-style model |
| Vocoder generator | **1,167,077** | Snake V2Mid HiFi-GAN-style generator |
| Total inference model | **4,632,202** | Acoustic + vocoder |

Model files:

```text
weights/inflect_nano_v1_acoustic.pt
weights/inflect_nano_v1_vocoder.pt
```

## Architecture

Inflect-Nano-v1 is a two-part TTS stack:

```text
Text
  -> English normalization + G2P frontend
  -> Compact FastSpeech-style acoustic model
  -> 80-bin mel spectrogram
  -> Snake V2Mid HiFi-GAN-style vocoder
  -> 24 kHz waveform
```

### Acoustic Model

The acoustic model predicts duration, energy, brightness, and pitch, expands token states into frame states, and decodes 80-bin mel spectrograms.

Main config:

```json
{
  "hidden": 168,
  "encoder_layers": 5,
  "decoder_layers": 6,
  "decoder_ff_mult": 3,
  "kernel_size": 7,
  "speaker_dim": 64,
  "dropout": 0.08,
  "n_mels": 80,
  "sample_rate": 24000,
  "max_frames": 1400,
  "postnet_scale": 0.1,
  "use_frame_pitch": true,
  "abs_frame_bins": 512
}
```

Approximate acoustic split:

```text
total acoustic: 3.465M
encoder:        1.292M
decoder:        1.211M
postnet:        0.276M
local context:  0.226M
frame GRU:      0.128M
heads/embeds/projections: remainder
```

### Vocoder

The vocoder is a custom Snake-activation HiFi-GAN-style generator. Discriminators are training-only and are not included in inference parameter counts.

Main config:

```json
{
  "variant": "snake_v2mid",
  "sample_rate": 24000,
  "n_fft": 1024,
  "hop_size": 256,
  "win_size": 1024,
  "num_mels": 80,
  "fmax": 12000.0,
  "upsample_rates": [8, 8, 2, 2],
  "upsample_kernel_sizes": [16, 16, 4, 4],
  "upsample_initial_channel": 144,
  "resblock_kernel_sizes": [3, 7, 11],
  "resblock_dilation_sizes": [[1, 3, 5], [1, 3, 5], [1, 3, 5]],
  "activation": "snake"
}
```

## Intended Use

Good fits:

- Tiny local TTS experiments
- Offline assistant prototypes
- Embedded or low-resource speech demos
- Efficient inference research
- Browser/WASM-style exploration
- Baseline for sub-5M TTS work

Poor fits:

- Production narration
- Accessibility-critical output
- Voice cloning
- Multilingual TTS
- High-fidelity studio speech
- Long-form audiobook generation

## Limitations

This model is intentionally tiny and has clear quality limits:

- Unseen text can stumble or sound unstable.
- The voice can sound robotic, buzzy, or artifacted.
- Long or unusual prompts are less reliable.
- The vocoder is a major quality bottleneck.
- It is not a voice cloning model.
- It is not multilingual.
- It is not suitable for safety-critical or production accessibility use.

## Recommended Framing

Use this as:

> An experimental 4.63M-parameter English TTS model exploring the quality/size tradeoff for ultra-small local speech synthesis.

Do not present it as SOTA or production-quality.

## License

Apache-2.0.

This repository includes a small third-party English text frontend derived from TinyTTS-style code for tokenization/G2P compatibility. Its license is included as `TINY_TTS_LICENSE`.