Audio & Speech

Today's Stories

1. 1

   Anthropic clashes with Trump's White House, rejects Washington playbook

   The Trump administration has twice taken actions against Anthropic—labeling it a "supply chain risk" in April after the company refused Pentagon contract language, and imposing export controls on its Mythos and Fable AI models two weeks ago following discovery of a jailbreak. OpenAI, by contrast, announced it was withholding release of GPT-5.6 at the U.S. government's request on the same day those controls were relaxed. Anthropic, valued at $965 billion（約150兆円） and preparing for an IPO expected in the coming months, has refused the flattery, donations, and appointment of Trump allies that other tech giants (Meta, Amazon, Apple) and OpenAI have used to stay in the administration's favor. Trump administration officials have publicly attacked CEO Dario Amodei as a "liar" with a "God-complex" and an "ideological lunatic," and accused the company of "regulatory capture." Continued hostility could make it harder to sell public market investors on the stock listing and significantly hamper the company's ability to develop advanced AI models.

   Anthropic CEO Amodei reportedly called Trump "a feudal warlord" in a now-deleted Facebook post, and his sister and cofounder Daniela Amodei donated to Kamala Harris's campaign. Unlike OpenAI's policy chief Chris Lehane and cofounder Greg Brockman (the largest donor to Trump Super PAC MAGA Inc.), Anthropic has made no similar hires of Trump-aligned figures to its leadership.

2. 2

   Netflix uses AI-generated Gene Wilder voice for Wonka reality show

   Netflix is premiering Wonka's The Golden Ticket on September 23rd, a reality competition based on the fictional Wonka universe. The show's voiceover uses an AI-generated version of Gene Wilder's voice, created in partnership with AI audio company ElevenLabs and with consent from Wilder's family. This extends Netflix's pattern of using AI-generated celebrity voices for content—the company has previously recreated voices of Michael Caine and Stan Lee. For viewers, it means encountering synthetic versions of iconic figures in new productions, blurring the line between archival and synthetic media in mainstream entertainment.

   The two-part finale airs on September 30th. The show will feature 12 golden ticket winners and their chosen partners competing in a high-stakes social experiment, with one champion crowned by the end.

3. 3

   Reddit user struggles to replicate Pocket TTS model from research paper

   A developer attempting to implement Pocket TTS (a text-to-speech model from kyutai-labs) without access to the original training code is reporting significant difficulties. Despite training on smaller datasets (LJSpeech and LibriSpeech clean subset) with reduced model parameters, the model is failing to generate coherent speech even on text from its training set. The flow matching loss reached approximately 0.20 mse, but inference at epoch 2800 produced largely meaningless output; attempts to fix this via scheduled sampling and noise injection did not resolve the problem. This illustrates a common friction point in AI research: papers are published without accompanying training or fine-tuning code, forcing practitioners who want to learn from or build on the work to reverse-engineer implementations from scratch. When that reconstruction fails, it blocks both reproducibility and the ability for others to build downstream applications, limiting the practical reach of published research.

   The core issue remains unresolved—the user has not identified the root cause of the model's failure to generate meaningful output despite low loss metrics. This gap between training loss and inference quality suggests a possible mismatch in model architecture, data preprocessing, or inference logic that would require either access to the original code or additional debugging from the community.

4. 4

   AI Voice Narration Still Lacks Emotional Depth, User Seeks Better Alternative

   A user working on animation narration experimented with Elevenlabs for AI voice generation but found the output felt flat and character voices were inconsistent. They also tested voice reference features with a 15-second limit and ultimately paid $75 for a human voice actor, whose work proved unusable. This highlights a real gap in current AI voice technology—existing solutions like Elevenlabs struggle to capture the emotional tone and character consistency that creative projects require. For animators and media producers, this means AI narration tools may still fall short of production quality, forcing fallback to costlier human talent or workarounds.

   The user is asking whether any AI voice tool currently outperforms Elevenlabs at emotional expression and consistency. The answer suggests no clear successor yet exists, leaving a potential opportunity for better-performing voice AI products in the market.

5. 5

   NagaTranslate: AI translation and voice tool for India's low-resource Naga languages

   A developer has built NagaTranslate, a project combining translation and speech processing for Nagaland's low-resource languages, currently supporting Nagamese, Ao, and Sema. The text translation backend uses a commercial LLM API with optimized prompts and few-shot examples; the developer initially tried a fine-tuned NLLB (No Language Left Behind) model before switching to the LLM approach. Nagamese and other native Naga languages were primarily oral with very little standard parallel data, making this an underserved gap in natural language processing. Building tools for these languages may help preserve and digitize speech and text in communities where written standards have only recently emerged.

   The project is open to feedback on its architecture and strategies for improving the pipeline under strict resource constraints—a challenge relevant to anyone working on translation or voice tools for languages with limited training data.

6. 6

   Which AI Voice Agent Stack Has the Lowest Latency?

   Which AI Voice Agent Stack Has the Lowest Latency?