Studio
Password
Sign in
Studio
Image
Video
Music
Voice
History
Mode
Generate
Edit / references
Upscale
Model
Prompt
Images (first is edited; up to 10 references)
Paint the area to change (mask)
Clear mask
Brush
Size
Auto
1:1
16:9
9:16
4:3
3:4
3:2
2:3
Quality
Auto (1024)
Medium (1536)
High (2048)
Count
Seed
Upscale factor
2
3
4
Transparent background (Qwen models)
Rewrite prompt with the model's own rewriter
Create
Model
Length (s)
Size
Auto (768p)
1344×768 (16:9)
768×1344 (9:16)
1024×1024
960×544 (faster)
544×960 (faster)
Seed
Prompt (picture, motion, sound and music)
First frame
Last frame
References instead of keyframes: images (≤9), videos (≤3), audio (≤3) — call them <Picture 1>, <Video 1>, <Audio 1> in the prompt
H3 renders 5–15 s with sound. On this hardware a 5 s 768p clip takes ~15 min with Turbo and about an hour at full quality.
Create video
Model
ACE-Step XL (best, all tasks, Finnish)
YuE2 (top vocals, scores)
ACE-Step turbo (fast)
Task
Generate
Cover
Repaint a section
Add an instrument track
Extract a stem
Complete with tracks
Format
mp3
flac
wav
Style / description
Lyrics ([Verse], [Chorus], …)
Language
Duration (s)
BPM
Key
Count
Seed
Instrumental
Let ACE-Step write the lyrics from the description
Source audio (cover / repaint / tracks)
Cover strength (0–1)
From (s)
To (s)
Track
—
vocals
backing_vocals
drums
bass
guitar
keyboard
percussion
strings
synth
fx
brass
woodwinds
Tracks (complete)
Style reference audio (ACE-Step)
Score in ABC (YuE2 "notes"; leave empty to let YuE2 compose)
Score mode
full (melody + chords)
melody
off
Create music
Audio → score
SheetSage2 transcribes a recording into ABC notes you can edit and give YuE2.
Audio
Mode
Melody
Melody + chords
Transcribe
Text to speech
Model
Voice
Language
Text
Speak
Speech to text
Model
Language
Audio file
● Record
Transcribe
All
image
video
music
speech
transcript
Refresh
Recent