[Speechify] Add support for Speechify as a TTS provider

[DaemonLoki]

What's implemented?

• Speechify as a TTS provider, usable with speechify.TTS()
• We're supporting different models, up to their latest simba-3.2
• We have support for the different voice (e.g. geffen_32 as the default, but others, like beatrice_32, etc)
• Multilingual support works as well, using e.g. speechify.TTS(model="simba-multilingual", language="de-DE")

Demo

speechify-demo.mp4

[coderabbitai]

📝 Walkthrough

Walkthrough

Adds a Speechify TTS plugin with environment-based API-key configuration, configurable voice/model/language settings, and streaming 24 kHz mono S16 PCM output. Registers the plugin in workspace and optional dependency configuration, documents its API, adds unit and integration tests, and provides a voice-bot example plus a standalone WAV-producing smoke test.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 1

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 34555113-4b2e-4f30-8d0e-40eabb37b338

📥 Commits

Reviewing files that changed from the base of the PR and between 019edac and b277fbb.

⛔ Files ignored due to path filters (1)

• uv.lock is excluded by !**/*.lock

📒 Files selected for processing (15)

• CHANGELOG.md
• agents-core/pyproject.toml
• plugins/speechify/README.md
• plugins/speechify/example/.env.example
• plugins/speechify/example/README.md
• plugins/speechify/example/main.py
• plugins/speechify/example/pyproject.toml
• plugins/speechify/example/tts_smoke.py
• plugins/speechify/py.typed
• plugins/speechify/pyproject.toml
• plugins/speechify/tests/__init__.py
• plugins/speechify/tests/test_tts.py
• plugins/speechify/vision_agents/plugins/speechify/__init__.py
• plugins/speechify/vision_agents/plugins/speechify/tts.py
• pyproject.toml

[coderabbitai]

🎯 Functional Correctness | 🔴 Critical | ⚡ Quick win

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
pip install speechify-api
python -c "
import inspect
from speechify import AsyncSpeechify
print('Is coroutine:', inspect.iscoroutinefunction(AsyncSpeechify().audio.stream))
"

Repository: GetStream/Vision-Agents

Length of output: 2237

---

🏁 Script executed:

#!/bin/bash
python3 - <<'PY'
import inspect
import speechify
from speechify import AsyncSpeechify

print("speechify module:", speechify.__file__)
print("AsyncSpeechify class:", AsyncSpeechify)

# Try to inspect the class attributes without instantiating a client.
audio_attr = getattr(AsyncSpeechify, "audio", None)
print("AsyncSpeechify.audio on class:", type(audio_attr), audio_attr)

# Locate source files for the package.
import speechify.client as client_mod
print("client module:", client_mod.__file__)

# Print relevant lines around AsyncSpeechify and any stream method definitions.
import pathlib, re
paths = [
    pathlib.Path(client_mod.__file__),
]
for path in paths:
    text = path.read_text()
    for needle in ["class AsyncSpeechify", "async def stream", "def stream"]:
        if needle in text:
            print(f"FOUND {needle} in {path}")
PY

Repository: GetStream/Vision-Agents

Length of output: 589

---

🏁 Script executed:

#!/bin/bash
python3 - <<'PY'
import pathlib
import inspect

import speechify.client as client_mod

path = pathlib.Path(client_mod.__file__)
text = path.read_text().splitlines()

# Show the AsyncSpeechify class and nearby stream-related methods.
needles = ("class AsyncSpeechify", "async def stream", "def stream")
for i, line in enumerate(text, start=1):
    if any(n in line for n in needles):
        start = max(1, i - 8)
        end = min(len(text), i + 20)
        print(f"\n--- {path}:{start}-{end} ---")
        for j in range(start, end + 1):
            print(f"{j:4d}: {text[j-1]}")
PY

Repository: GetStream/Vision-Agents

Length of output: 1461

---

🏁 Script executed:

#!/bin/bash
python3 - <<'PY'
from pathlib import Path
import speechify.client as client_mod

path = Path(client_mod.__file__)
print(path)
for i, line in enumerate(path.read_text().splitlines(), start=1):
    if "async def stream" in line or "def stream" in line or "class AsyncSpeechify" in line:
        print(f"{i}: {line}")
PY

Repository: GetStream/Vision-Agents

Length of output: 259

---

🏁 Script executed:

#!/bin/bash
python3 - <<'PY'
from pathlib import Path
import speechify
import speechify.client as client_mod

paths = [Path(speechify.__file__), Path(client_mod.__file__)]
for path in paths:
    print(f"\nFILE: {path}")
    text = path.read_text().splitlines()
    for i, line in enumerate(text, start=1):
        if "class AsyncSpeechify" in line or "async def stream" in line or "def stream" in line:
            start = max(1, i - 6)
            end = min(len(text), i + 18)
            for j in range(start, end + 1):
                print(f"{j:4d}: {text[j-1]}")
            print()
PY

Repository: GetStream/Vision-Agents

Length of output: 1213

---

🌐 Web query:

speechify-api Python SDK AsyncSpeechify audio stream await

💡 Result:

To perform asynchronous audio streaming using the Speechify Python SDK, you should use the AsyncSpeechify client class [1][2][3]. The SDK supports non-blocking operations, allowing you to iterate over audio chunks as they are generated by the API [4]. ### Async Streaming Implementation The SDK provides an asynchronous interface for streaming. You can use an async for loop to process the audio chunks returned by the client.audio.stream method [4]. Example usage: python import asyncio from speechify import AsyncSpeechify async def stream_audio(): # Initialize the async client client = AsyncSpeechify(token="YOUR_API_KEY") # Open the stream stream = await client.audio.stream( accept="audio/mpeg", input="Hello! This is a streaming test using the Speechify Python SDK.", voice_id="george" ) # Iterate over incoming audio chunks async for chunk in stream: # Process each binary chunk (e.g., write to a file or send to a buffer) print(f"Received chunk of size: {len(chunk)} bytes") # Run the async function asyncio.run(stream_audio()) ### Key Considerations * Async Client: Always instantiate AsyncSpeechify instead of the synchronous Speechify client to access non-blocking methods [1][2]. * Custom HTTP Client: If you need advanced control (e.g., custom timeouts or proxy settings), you can pass a httpx.AsyncClient() instance to the httpx_client parameter during the initialization of AsyncSpeechify [1][4]. * Resource Management: Ensure you are using asyncio or a compatible event loop runner to execute these calls [1][2]. * Cookbook Resources: For more complex patterns, including voice cloning or saving streams directly to disk, refer to the official Speechify API Cookbook on GitHub [5].

Citations:

• 1: https://pypi.org/project/speechify-api/
• 2: https://github.com/SpeechifyInc/speechify-api-sdk-python
• 3: https://pypi.org/project/speechify-api/3.0.1/
• 4: https://context7.com/speechifyinc/speechify-api-sdk-python/llms.txt
• 5: https://github.com/SpeechifyInc/speechify-api-cookbook

---

Await self.client.audio.stream(...). audio_stream is a coroutine here, so PcmData.from_response(...) will fail.

[luke-speechify]

Suggested change

	"speechify-api>=1.2.4,<2",
	"speechify-api==3.0.1",

recommend jumping to 3.0.1 but i did break the wire when going to v3 (sorry!)

[luke-speechify]

Suggested change

	client if client is not None else AsyncSpeechify(api_key=self.api_key)
	client if client is not None else AsyncSpeechify(token=self.api_key)

i believe this is the breaking bit (again, sorry!)