Skip to main content

AIGCPanel v2.5.0: Cloud Voice Models Now on Desktop, No GPU Required

AIGCPanel v2.5.0: Cloud Voice Models Now on Desktop, No GPU Required

If you've ever stared at your modest laptop and sighed because it can't handle AI voice synthesis, AIGCPanel's latest update is for you. Version 2.5.0, released today, brings cloud-based voice capabilities straight to your desktop—no graphics card required.

The magic happens on the AI model page. Under "Other Models," you can now enter an API key for Volcano Engine's Dobao Voice or Alibaba Cloud's DashScope. That's it. The software instantly activates the cloud voice model, letting you generate high-quality speech without local hardware.

A Smarter Way to Add Voices

But wait—there's more. Before saving your key, AIGCPanel runs a quick test: it synthesizes a simple phrase, "Hello, this is a voice preview." If the connection works, you can save; if not, you'll know right away. No more digging through error logs later.

Behind the scenes, the main process maintains a registry of voice providers. Each vendor just needs to adapt its HTTP interface to a unified soundTts call. Dobao Voice uses X-Api-Key authentication and returns a stream of JSON segments that the software merges into a complete audio file. Alibaba Cloud's DashScope qwen-tts returns an audio link or base64 data. Adding a new vendor in the future? Just write one provider file.

Separate Libraries for Digital Humans and Live Streaming

This update also centralizes voice management. On the digital human side, you have "voice tones"; on the live streaming side, "live streaming tones." Each has its own library, and both support adding synthesized or cloned voices with a preview before saving. The two libraries are stored separately, so they don't interfere.

For voice cloning, you need to re-record or re-upload reference audio along with corresponding text—no more reusing the digital human's voice library. That means no more confusion about which user influenced a cloned voice. The add dialog also includes a handy guide on recording requirements and operation tips.

At the core is a VoiceService factory with two storage identifiers: SoundVoice and LiveVoice. Each voice record stores type, model identifier, parameters, and reference audio path. Real-time synthesis for preview and live streaming uses the same synthesize() call chain, ensuring consistent results. Even better, changing voices during a live stream doesn't require modifying the live streaming service—the client just exposes a provider name with a Voice: prefix, and routing is handled internally. Adding a new voice in the future takes just one line of code on the live streaming end. And don't worry, old configurations still work.

Live Streaming Control Panel Gets an Upgrade

The live streaming control panel now has a "Model Control" area at the top, where you can start or stop the live model and check its status. Next to it, a "Model Log" button opens real-time logs. If the live model hasn't been imported yet, it prompts you and jumps to the AI model interface. Previously, you had to switch between pages to confirm if the model was running; now it's all in one place. This was built using existing components—filtering the model list for live-capable records and reusing status badges and log viewers.

Polished Interface and Bug Fixes

Interface details have been refined: dialogs expand naturally without overflowing small windows; the selected item in the left navigation now has a light blue brand shade, with dark mode adjusted for brightness; disabled buttons use a light gray background with medium gray text for better readability. Several bugs have been squashed: general model tasks were mistakenly marked successful on abnormal exits—now they're verified for actual output; some error messages still appeared in Chinese on English interfaces; and errors from executing wmic on newer Windows versions are gone.

Key Points

  • Cloud voice without GPU: Enter API keys for Dobao Voice or DashScope to activate cloud-based TTS.
  • Separate voice libraries: Digital human and live streaming tones are managed independently.
  • Streamlined live streaming: New model control panel and real-time logs.
  • Interface improvements: Better dialogs, navigation, and button readability.
  • Bug fixes: Accurate task status, localized errors, and Windows compatibility.

Image