Upload Audio + Character
Audio is required for transcription. Character reference is required before image generation so identity stays locked.
Agent SDK workflow
Lyric Video Forge
Build ILLCO-ready lyric video runs from uploaded audio, uploaded character references, Realtime-first transcription, user-approved lyrics, selected image counts, dissolve visuals, readable ASS subtitles, strict rhyme coloring, and FFmpeg QC notes.
Agent SDK production briefCredits usage meteringUpload audio + referenceChatGPT App MCP endpoint
Default laneM3ntally-iLL lyric videos with timing discipline
Transcribe first, confirm lyrics, generate only the credit-selected image count, then cross-dissolve the visuals. No fake exact-sync claims.
Transcribe + Confirm Lyrics
The app uses a rap-specialist transcription prompt and blocks image generation until the user confirms the lyrics are correct.
Generate Images + Dissolve
The selected image count controls credit use. Generated stills are dissolved together instead of hard-cut.
Forge A Lyric Video Run
Upload audio, upload a character reference, transcribe first, approve the lyrics, then generate the selected number of visuals and dissolve them.
Credit system
Lyric Video Forge Credits
Credits meter transcription, Agent SDK planning, image generation count, timing-plan passes, and full render-plan runs.
1 credits current action22 image credits selectedAgent: gpt-5-nano / Realtime: gpt-4o-transcribe / Word timing: whisper-1 / Images: gpt-image-1
ChatGPT App endpoints
Connect this Forge
MCP https://illco-command.vercel.app/api/chatgpt/lyric-video-forge/mcpSSE https://illco-command.vercel.app/api/chatgpt/lyric-video-forge/sseWidget https://illco-command.vercel.app/api/chatgpt/lyric-video-forge/widget
Setup endpoints