skills.rest
zamansepeti43
zamansepeti43COMMUNITY
Explore all skills by this author →
GitHub

hyperframes-media

Generates TTS voiceovers, background music, sound effects, transcriptions, and captions for HyperFrames video compositions.

Updated Sep 1, 2026
Search Tags:#captions#transcription#whisper#sound effects#video production#tts#background music
One-click install
npx skills add https://github.com/zamansepeti43/c-rak-agent --skill hyperframes-media-zamansepeti43
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: hyperframes-media
Source: https://github.com/zamansepeti43/c-rak-agent/tree/main/video-engine/.agents/skills/hyperframes-media
Command: npx skills add https://github.com/zamansepeti43/c-rak-agent --skill hyperframes-media-zamansepeti43

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires google-genai, kokoro-onnx, soundfile, transformers, torch, numpy, elevenlabs, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve? Producing the audio layer of a video — voiceover, background music, sound effects, and timed captions — normally requires juggling multiple TTS providers, music generators, transcription tools, and caption formats. This Skill centralizes all of that into one audio engine that picks the right provider automatically and degrades gracefully when credentials are missing. ## Core Features & Use Cases - Multi-provider TTS: Synthesizes voiceovers via HeyGen (with native word timestamps), ElevenLabs, or local Kokoro, selected by credential availability. - BGM and SFX: Retrieves music and sound effects from the HeyGen audio library when credentialed, or falls back to local Lyria/MusicGen generation and a bundled 21-file SFX library offline. - Transcription and captions: Produces word-level timestamps via Whisper and authors styled, animated captions (karaoke, per-word emphasis, marker effects) consumed by HTML compositions. - Use Case: Given a storyboard with narration lines and mood cues, write an audio_request.json, run the engine, and receive an audio_meta.json with voice files, word timings, a BGM track, and placed SFX cues ready for assembly. ## Quick Start Ask the agent to generate a voiceover, background music, and captions for your video script using the hyperframes-media audio engine.

Frequently Asked Questions about hyperframes-media

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate a voiceover with word-level timestamps for captions?▼

Use HeyGen TTS, which returns native word timestamps in the same response as the audio — pass --words to capture them directly. For ElevenLabs or Kokoro, which return no word data, chain a Whisper transcribe pass on the generated audio to get the same flat word array.

What TTS providers are supported and how is one selected?▼

Three providers are chained in priority order: HeyGen (Starfish) when a credential resolves, then ElevenLabs when its API key is set, then local Kokoro-82M as the always-available fallback. You can pin a provider explicitly, but voice IDs are provider-specific.

Can I generate background music without a HeyGen API key?▼

Yes. Without a HeyGen credential the engine falls back to local generation: Google Lyria if a Gemini/Google API key is present, otherwise MusicGen (facebook/musicgen-small) running locally. Generation runs detached, so run wait-bgm.mjs before assembling the video.

Why does Whisper transcription translate my non-English audio into English?▼

The CLI default model is small.en, and .en models silently translate non-English audio into English. Always pass --model explicitly: use small.en for known English, or small --language <code> for known non-English audio.

What happens if a sound effect or music track cannot be found?▼

Missing audio never blocks a render. An unmatched SFX cue or BGM query is skipped and recorded as a non-fatal anomaly in the engine output, while voice and other assets still render normally.

When is remove-background the wrong tool for isolating a subject?▼

The --background-output plate is hole-cut, not inpainted — the subject region is fully transparent. If you need the room shown without the person as a standalone image, you need an inpainter such as LaMa, ProPainter, or E2FGVI instead.

Related Skills

Explore More Design & Creative Skills→
mmcmediammcmedia
2

creative-director

Analyze songs and generate scene-by-scene JSON storyboards for music videos.

Community
Advanced
X-manistX-manist

hyperframes-media

Generate TTS voiceovers, background music, sound effects, and captions for HyperFrames video compositions.

Community
Advanced
guanyangguanyang
960

remotion-captions

Transcribe, display, and animate captions in Remotion videos using @remotion/captions.

Community
Intermediate
p-possp-poss
26

export

Exports active Strudel mix code with timestamp and layer name header comments.

Community
Basic