Gandr TTS in Dify.
One plugin install inside Dify, and your workflows and agent apps can speak their output.
Ask Claude Code
Ask ChatGPTThe whole integration
Install the plugin, paste your key.
# Inside Dify: Plugins, then Install plugin.
# The plugin is published at github.com/Gandr-AI/dify-plugin,
# installable by GitHub repo or by plugin file.
# Credentials: your gnd_ API key, set once for the workspace.
The plugin adds Gandr text-to-speech as a tool node, so a workflow or agent app produces spoken audio as part of its run.
What it needs
What it takes
| Dify | A workspace with plugins enabled |
| One plugin | Gandr, from github.com/Gandr-AI/dify-plugin |
| A key | Your gnd_ API key, pasted into the plugin credentials |
Speech as a workflow step
Dify apps are graphs of model calls, tools and logic. The Gandr plugin registers text-to-speech as a tool, so a node in the graph takes text in and returns spoken audio like any other tool call.
The plugin talks to the same endpoint as every other integration here: an OpenAI-compatible speech contract with streaming support, fronted by one key.
What the flat stream changes
A Dify app that answers in text pays nothing extra to answer out loud on a flat stream. The workflow talks as much as the runs need and nothing is counted per character.
- A tool node, not a sidecar service
- Same OpenAI-compatible contract as the REST integration
- One stream carries the whole workspace workload
Notes
Where do I get the plugin?
From github.com/Gandr-AI/dify-plugin, installed into Dify by repository or by plugin file.
Which voices does the plugin expose?
The stock gandr- voices, and cloned voices through the same API the plugin uses.