--- title: "WaveRNN vs dia2" type: "comparison" canonical_url: "https://www.graphcanon.com/compare/fatchord-wavernn-vs-nari-labs-dia2" tools: ["fatchord-wavernn", "nari-labs-dia2"] --- # WaveRNN vs dia2 *GraphCanon updated Sep 20, 2026* ## Verdict Pick WaveRNN when license: WaveRNN is MIT, dia2 is Apache-2.0; pick dia2 when license: dia2 is Apache-2.0, WaveRNN is MIT. [WaveRNN](https://fatchord.github.io/model_outputs/) reports 2.2k GitHub stars, 685 forks, and 108 open issues, last pushed Jul 2, 2022. [dia2](https://github.com/nari-labs/dia2) has 1.2k stars, 102 forks, and 5 open issues, last pushed Nov 29, 2025. Figures are from public GitHub metadata via [WaveRNN's repository](https://github.com/fatchord/WaveRNN) and [dia2's repository](https://github.com/nari-labs/dia2). | | [WaveRNN](/tools/fatchord-wavernn.md) | [dia2](/tools/nari-labs-dia2.md) | | --- | --- | --- | | Tagline | WaveRNN Vocoder + TTS | TTS model capable of streaming conversational audio in real-time. | | Stars | 2,191 | 1,180 | | Forks | 685 | 102 | | Open issues | 108 | 5 | | Language | Python | Python | | Adopt for | WaveRNN is a Python-based neural vocoder that can generate high-quality speech from text when used with TTS models like Tacotron. | - | | Persona | - | - | | Runtime | - | - | | License | MIT | Apache-2.0 | | Categories | Speech & Audio | Speech & Audio | ## Trust and health _Sourced signals - not a safety guarantee. No winner column._ | | [WaveRNN](/tools/fatchord-wavernn.md) | [dia2](/tools/nari-labs-dia2.md) | | --- | --- | --- | | Maintenance | Dormant (18%) | Slowing (36%) | | Days since push | 1540d | 295d | | Open issues (now) | 108 | 5 | | Stars delta | +1 (30d) | +20 (30d) | | Owner type | User | Organization | | Full report | [trust report](/tools/fatchord-wavernn/trust.md) | [trust report](/tools/nari-labs-dia2/trust.md) | ## Decision facts: WaveRNN - **Requirements:** Python version must be equal to or higher than 3.6; Pytorch 1 with CUDA support is a prerequisite - **Adopt for:** WaveRNN is a Python-based neural vocoder that can generate high-quality speech from text when used with TTS models like Tacotron. ## Choose when ### Choose WaveRNN if… - License: WaveRNN is MIT, dia2 is Apache-2.0. - Requirements: Python version must be equal to or higher than 3.6; Pytorch 1 with CUDA support is a prerequisite. - Tags unique to WaveRNN: neural-vocoder, pytorch, speech-synthesis, tacotron. - When you require a compact and efficient method to produce natural-sounding speech synthesis, specifically with the need for high fidelity in audio quality. ### Choose dia2 if… - License: dia2 is Apache-2.0, WaveRNN is MIT. - Tags unique to dia2: open-weight. - When you need real-time streaming capabilities for text-to-speech applications. ## When NOT to use WaveRNN - Avoid using when you need extensive customization of the vocoder parameters, since it is optimized for specific configurations and might not offer the level of tweakability other frameworks provide. - Not recommended if your setup does not support CUDA, as WaveRNN requires PyTorch with CUDA for execution. ## When NOT to use dia2 - For projects aimed at batch processing or non-realtime TTS applications where instantaneous audio output isn't required. - If you are looking for an open-source model that does not include specific proprietary components like the Kyutai Mimi codec. ## Common questions ### What is the difference between WaveRNN and dia2? WaveRNN: WaveRNN Vocoder + TTS. dia2: TTS model capable of streaming conversational audio in real-time.. See the comparison table for live GitHub stats and shared categories. ### When should I choose WaveRNN over dia2? Choose WaveRNN over dia2 when License: WaveRNN is MIT, dia2 is Apache-2.0; Requirements: Python version must be equal to or higher than 3.6; Pytorch 1 with CUDA support is a prerequisite; Tags unique to WaveRNN: neural-vocoder, pytorch, speech-synthesis, tacotron; When you require a compact and efficient method to produce natural-sounding speech synthesis, specifically with the need for high fidelity in audio quality. ### When should I choose dia2 over WaveRNN? Choose dia2 over WaveRNN when License: dia2 is Apache-2.0, WaveRNN is MIT; Tags unique to dia2: open-weight; When you need real-time streaming capabilities for text-to-speech applications. ### When should I avoid WaveRNN? Avoid using when you need extensive customization of the vocoder parameters, since it is optimized for specific configurations and might not offer the level of tweakability other frameworks provide. Not recommended if your setup does not support CUDA, as WaveRNN requires PyTorch with CUDA for execution. ### When should I avoid dia2? For projects aimed at batch processing or non-realtime TTS applications where instantaneous audio output isn't required. If you are looking for an open-source model that does not include specific proprietary components like the Kyutai Mimi codec. ### Is WaveRNN or dia2 more popular on GitHub? WaveRNN has more GitHub stars (2,191 vs 1,180). Stars measure visibility, not whether either tool fits your constraints. ### Are WaveRNN and dia2 open source? Yes - both are open-source projects on GitHub (WaveRNN: MIT, dia2: Apache-2.0). ### Where can I find alternatives to WaveRNN or dia2? GraphCanon lists graph-backed alternatives at [WaveRNN alternatives](/tools/fatchord-wavernn/alternatives) and [dia2 alternatives](/tools/nari-labs-dia2/alternatives) ([WaveRNN markdown twin](/tools/fatchord-wavernn/alternatives.md), [dia2 markdown twin](/tools/nari-labs-dia2/alternatives.md)), ranked by typed relationship edges rather than popularity votes. ### Is there a machine-readable version of this comparison? Yes. The markdown twin at [this comparison](/compare/fatchord-wavernn-vs-nari-labs-dia2.md) mirrors this page for agents and LLM crawlers, with the same stats table and FAQ answers. ### Which is better maintained, WaveRNN or dia2? WaveRNN: Dormant. dia2: Slowing. Compare maintenance labels, days since push, and release cadence in the trust section below - stars alone do not measure maintenance. ### Where are the full trust reports for WaveRNN and dia2? GraphCanon publishes per-repo trust reports with dated maintenance, provenance, and scan summaries: [WaveRNN trust report](/tools/fatchord-wavernn/trust); [dia2 trust report](/tools/nari-labs-dia2/trust). --- **Machine-readable endpoints** - JSON: [`/api/graphcanon/graph?tool=fatchord-wavernn`](/api/graphcanon/graph?tool=fatchord-wavernn) - LLM index: [/llms.txt](/llms.txt) - Full corpus: [/llms-full.txt](/llms-full.txt) _GraphCanon - The knowledge graph for AI development. https://www.graphcanon.com/_