v0.11.0

v0.11.0 (2026-10-24)

Breaking changes

• heygen model audio voices create no longer takes --name, --language, --mode or --voice-id. The endpoint now accepts either a professional or an instant voice request, so pass those fields in the body instead, e.g. -d '{"mode":"professional","name":"Narrator","audio":[...]}'. Scripts that pass them as flags now fail with an unknown-flag error. (#512)

New

• Speech-to-speech. heygen model audio sts create re-speaks a recording of up to 300 seconds in one of your instant voices, keeping its words and timing. (#512)
• Identity-preserving image generation. heygen model images create generates one to five portraits that keep a person's identity while following a reference image's pose, clothing and background, and heygen model images get polls the result. Available to workspaces in the private beta. (#512)
• Instant voices. heygen model audio voices create can now create an instant voice from a single recording with "mode":"instant". (#512)
• More tts create options. heygen model audio tts create adds --model (including the lower-latency turbo models), --with-timestamps for character and word timings, --expressiveness-boost, and --end-user-id. --language is now optional and is detected from the text when omitted. (#512)
• Public voices. heygen model audio voices list --type public lists the voices HeyGen provides to every workspace. (#512)
• New option values. heygen voice speech create --engine accepts heygen_voice, and the webhook commands accept the video_generation.success and video_generation.fail events. (#512)

Internal

• Built with Go 1.26.10, which fixes several Go standard-library security advisories (#513)
• Codegen and release docs updates (#456, #435)