| --- |
| title: Plotweaver Nigerian Voice |
| emoji: 🗣️ |
| colorFrom: green |
| colorTo: blue |
| sdk: gradio |
| sdk_version: 6.19.0 |
| python_version: "3.10" |
| app_file: app.py |
| pinned: false |
| license: cc-by-nc-4.0 |
| --- |
| |
| # Plotweaver Nigerian Voice |
|
|
| Nigerian-English text-to-speech, fine-tuned from [F5-TTS](https://github.com/SWivid/F5-TTS) |
| by Plotweaver AI. |
|
|
| Two modes: |
| - **Nigerian Voice** — type text and hear it in the fine-tuned Nigerian-English voice. |
| - **Clone a Voice** — upload a short reference clip and generate speech in that voice |
| (transcript optional; auto-transcribed with Whisper when left blank). |
|
|
| Runs on free CPU hardware, so generation takes ~30–90+ seconds per request. |
| Research / non-commercial use only (base F5-TTS is CC-BY-NC-4.0). |
|
|
| ## Configuration notes |
| - `python_version: "3.10"` is pinned so `torch==2.4.0` (the tested training/inference |
| version) resolves. On Python 3.13 that wheel doesn't exist — bump torch to 2.5.1. |
| - `sdk_version: 6.19.0` controls Gradio; do not also pin `gradio` in `requirements.txt`. |
| - `HF_TOKEN` must be set as a Space **secret** with read access to the private model |
| repo `PlotweaverAI/nigerian-english-ft-tts`. |
| - Optional `F5TTS_ARCH` variable (defaults to `F5TTS_v1_Base`) if the checkpoint was |
| fine-tuned from a different base. |
|
|
