StepFun's whole StepAudio 3 voice lineup is free on its API until November 8: realtime calls, voice cloning and AI music
StepFun is letting developers use its StepAudio 3 preview models for free on its API until 23:59:59 on November 8, 2026 (UTC+8), which is 8:59 PM PKT. The preview lineup covers realtime voice conversation (stepaudio-3-realtime-preview), audio chat (stepaudio-3-chat-preview), speech and voice generation including cloning and voice design (stepaudio-3-gen-preview), and music (stepaudio-3-music-preview). Paid pricing starts November 9, and the preview model IDs are retired after the trial. Claude Code or any coding agent can wire them into a prototype quickly.
Why it matters
Realtime voice and cloning are usually billed per minute or per voice, which makes testing expensive. Four days of free access is enough to compare StepAudio 3 against the voice stack you already use.
Who it's for
Developers building voice agents, phone bots, narration or music features who want to benchmark a new voice model for free.
How to try it
- Create a StepFun platform account and get an API key.
- Ask Claude Code to build a quick test with stepaudio-3-realtime-preview or stepaudio-3-gen-preview, reading the key from an environment variable.
- Run your real scripts or voice samples through it before 8:59 PM PKT on November 8.
- If you keep it, switch to the non-preview model IDs and budget for the paid rates from November 9.
Plan & availability: Free on the StepFun API until 23:59:59 November 8, 2026 (UTC+8), 8:59 PM PKT. From November 9, StepFun lists realtime at $0.125 per session minute, TTS at $14 per million characters, voice cloning or design at $1.50 per voice, music at $2.40 to $4.80 per output hour and ASR at up to $0.60 per audio hour. The -preview model IDs stop working after the trial. Checked on StepFun's pricing page on November 5, 2026.
Related
Source: @StepFun on platform.stepfun.ai ↗
Written by AgentStackList from the original post. Checked Nov 5, 2026. Features, plans and availability can change, so confirm on the source before you rely on it.
freestepfunvoicetext-to-speechrealtime
More Claude tricks
- Fish Audio's top voice model is $0 until November 30, voice cloning and 83 languages included · Nov 5
- Tencent's new open model Hy4 is free in its CodeBuddy and WorkBuddy agents until November 15, and half price on its API · Nov 5
- $100 of speech-to-text and text-to-speech, no card: Speechmatics gives new sign-ups a credit that lasts 12 months · Nov 5
- Alibaba open-sourced qz, a browser agent that works inside the Chrome you're already logged into, and Claude Code can hand it tasks · Nov 5
Agent Stack Weekly · free
Get the week's best AI tricks
Limit and pricing changes, new MCP servers, and setup fixes for Claude Code, Codex, Cursor, Copilot and Antigravity.
No spam. Unsubscribe anytime. Read past issues