@gkxspace: I spend two to three thousand on AI subscriptions every month, some for TTS, ASR, etc. The mainstream ones are expensive and their API protocols differ. I kept thinking: is there a single plan that covers voice cloning, meeting transcription, AI podcast generation, real-time voice Q&A, voice input, and coding? Finally found a godsend—StepFun's S...
Summary
StepFun launches Step Plan subscription at $6.99/month, integrating LLM, TTS, ASR, image generation, and other AI models. Supports direct OpenAI SDK connection, applicable for voice cloning, meeting transcription, AI podcast generation, etc.
View Cached Full Text
Cached at: 05/20/26, 04:35 PM
I used to spend two to three thousand yuan a month on AI subscriptions, some of which were for TTS, ASR, etc. The mainstream services are quite expensive, and their API protocols are all different.
I’ve always been looking for a single plan that could do it all: voice cloning, meeting transcription, AI podcast generation, real-time voice Q&A, voice input, and code writing.
Finally found a true lifesaver — Step Plan by StepFun. It costs $6.99 per month and I can never use it all up. So I gradually canceled all the others.
One subscription gets you access to top-tier models of all kinds:
- LLM: Step 3.5 Flash — incredibly low latency, and you can also integrate it with Claude / Cursor / Cline.
- TTS: stepaudio-2.5-tts (I checked; its ranking is higher than ElevenLabs).
- ASR: Real-time voice conversations with voice cloning support.
- Image generation: Text-to-image + image editing, generating images in 0.7 seconds.
All accessible directly via the OpenAI SDK — just change the base URL.
Here are some use cases (details in the comments):
- English audio recording → Chinese notes in 54 seconds
- Long English article → dual‑speaker MP3 for commuting
- Same text → TTS with 7 different emotions
- Lu Xun’s Kong Yiji → automatic role‑based audiobook
- English podcast → end-to-end Chinese remake
@StepFun_ai
Similar Articles
@FinanceYF5: AI subscription plan subsidies are much larger than imagined. Claude Max 20x: $200/month, actual usage value about $8,000. ChatGPT Pro 20x: $200/month, actual usage value about $14,000. You spend $200, they lose thousands supporting you. This price war,…
Discusses the subsidy scale of AI subscription plans, pointing out that Claude Max and ChatGPT Pro cost $200 per month but actual usage value is far higher, implying fierce price competition.
@iluciddreaming: Don't rush to pay a monthly fee for AI avatars. Open source can already do it: a photo + an audio clip = a talking video. Free, self-hosted. That's essentially what platforms sell. Repo link in reply.
This tweet points out that open-source solutions can already turn a photo and an audio clip into a talking video, so there's no need to pay a monthly fee for AI avatar platforms. The repo link is in the reply.
@FeitengLi: Next week, after adding speaker labeling and speech generation, it won't be this cheap early bird price anymore.
EdgeSpeak officially launched, a local-first, privacy-preserving accurate transcription tool, supporting semantic segmentation and timestamps, compatible with OpenAI Audio API, etc. It will later add speaker labeling and speech generation features.
@AISuperDomain: Real-time voice agents are moving from "demo toys" to truly usable, and LiveKit Agents is arguably one of the most complete open-source frameworks. It doesn't just chain STT, LLM, and TTS; it also directly provides: • WebRTC real-time audio/video • Inbound and outbound phone calls • …
LiveKit Agents is an open-source real-time voice agent framework supporting WebRTC, telephony integration, semantic turn detection, MCP tool calling, and multi-agent handoff, helping developers quickly build real-time voice applications such as AI customer service and phone bots.
@MaxForAI: If you are working on voice agents, you should try this project. A team from NTU, NUS, and Shanghai AI Lab released: Mega-ASR. This fully open-source ASR is built on Qwen3-ASR, aiming to break the long-standing bottleneck of ASR performance in noisy, reverberant, or other impaired real-world environments...
NTU, NUS, and Shanghai AI Lab jointly released Mega-ASR, a fully open-source ASR model built on Qwen3-ASR. Using the Voices-in-the-Wild-2M dataset and progressive acoustic-to-semantic optimization, it achieves up to 30% relative Word Error Rate (WER) reduction in real-world noisy environments. With only 1.7B parameters, it enables efficient inference on consumer-grade hardware.