DEDUPED Hot AI repos. Real use cases. What people build. Get the Monday 5

How to run F5 TTS in a container instead of installing it with F5 TTS

Use the published image or build the Dockerfile so the CUDA and PyTorch setup is fixed.

Short answer: Build the image with docker build -t f5tts:v1 . from the repo, or pull the published one; then run the container from the registry with the README command for ghcr.io/swivid/f5-tts:main, mounting the f5-tts volume onto the Hugging Face hub cache and publishing port 7860; To go straight to the web interface append f5-tts_infer-gradio --host 0.0.0.0 to that run command. All steps.

  • Content
  • Intermediate
  • 1 hour
  • From the README
  • Claude Code
  • Codex
  • Cursor
  • Gemini CLI

platform. The documented run commands pass --gpus=all, so they expect an NVIDIA container runtime.

Who it is forA team that wants the same TTS setup on every machine.

The problemMatching CUDA, PyTorch and FFmpeg versions by hand breaks often.

The outcomeA container that runs the CLI or the web interface the same way everywhere.

What you need first

  • Python 3.10 or newer, for example a conda env created with `conda create -n f5-tts python=3.11`
  • PyTorch and torchaudio matched to your device, plus FFmpeg
  • A reference WAV clip of the voice you want, and its transcript if you do not want ASR to transcribe it
  • A GPU for comfortable speed; the published benchmarks use a single L20

Steps

  1. Build the image with docker build -t f5tts:v1 . from the repo, or pull the published one
  2. Run the container from the registry with the README command for ghcr.io/swivid/f5-tts:main, mounting the f5-tts volume onto the Hugging Face hub cache and publishing port 7860
  3. To go straight to the web interface append f5-tts_infer-gradio --host 0.0.0.0 to that run command
  4. For an NVIDIA machine copy the docker compose example in the README

Prompt for your agent

Using SWivid/F5-TTS, get the docker image running with the hub cache volume and port 7860, then generate one sample to prove the setup works.

Caveat. Model weights download into the mounted cache volume on first run.

Use this in your AI agent

Ask your agent

Using the Deduped MCP, get the setup steps for SWivid/F5-TTS and "Run F5 TTS in a container instead of installing it"

Add Deduped to Claude Code once

claude mcp add --transport http deduped https://dedupedai.com/mcp

Cursor, ChatGPT, others

Share: X LinkedIn Hacker News Email Copy link

Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer