draft 2026-10-19
last checked 2026-10-22
The soundtrack for your Claude music video: Suno or ismail?
Last checked: 22 October 2026. Plans and terms change; each line about Suno links to Suno's own page.
Short answer
If your video needs a sung song with lyrics, use Suno: ismail has no singing voice. If it needs an instrumental, a score or a beat you can edit bar by bar, made by the same agent that writes your visuals and timed to them exactly, use ismail: a free, open-source DAW your agent drives, running on your own computer.
The problem this solves
You asked Claude to build a music video in code. The visuals are code, so you can change any frame. The track came from somewhere else as one finished audio file: to change the drop, you generate again and hope; to sync a scene to a snare, you hunt for it by ear.
With ismail the music is code too. The agent writes the notes, so it knows when every kick, snare and chord change lands.
Side by side
| ismail | Suno | |
|---|---|---|
| What you get | A project your agent wrote: notes, instruments, effects, mix. Rendered to WAV, one stem per track | A finished, mixed song from a prompt, vocals included |
| Vocals | No singing voice. Speech and vocoder-style voices only | Yes, sung lyrics |
| Changing one part | Edit one note, bar, sound or fader, render again; everything else stays as it was | Regenerate or replace a section. Suno Studio (Premier plan) edits MIDI on a timeline (release notes) |
| Timing for your visuals | The notes are the source, so the agent knows when every note lands. python -m ismail.video sync writes every note with its video frame (to the nearest frame), plus loudness per frame, to a JSON file your video code can read | You get audio. Timing comes from listening, or from analysing the file |
| Files on your disk | Every render, as many as you like | Free: up to 7 downloads for the life of the account. Pro (8 USD a month): 20 a month. Premier (24 USD a month): 60 a month (pricing, downloads) |
| Using it in a published video | MIT licence, commercial use included. Some sampled or measured instruments ask for a credit, and the credits op lists them | Free plan: "No commercial rights" (pricing). Paid plans: commercial use of downloaded songs (terms) |
| Agent access | Built for agents: MCP tools, a command line and a Python API | No public API; the terms forbid "robots, scraping" (terms) |
| Runs | On your machine | In Suno's cloud |
| Setup | Your agent installs it | Nothing to install |
| Speed to a first full song | Minutes to hours, and it needs your direction | Under a minute |
What the timing file looks like
This is the start of the file ismail wrote for the music video of Crossroads Rewired (144 BPM, 30 frames a second). Each note carries its frame f, so the video code cuts on the kick and flashes on the snare without anyone listening for them:
{
"fps": 30,
"bpm": 144.0,
"frames_per_beat": 12.5,
"beats_per_bar": 4,
"length_bars": 162,
"total_frames": 8220,
"tracks": {
"kick": [
{"f": 1600, "beat": 128.0, "pitch": 36, "len": 3, "vel": 127},
{"f": 1631, "beat": 130.5, "pitch": 36, "len": 3, "vel": 100},
{"f": 1650, "beat": 132.0, "pitch": 36, "len": 3, "vel": 127}
],
"snare": [
{"f": 1625, "beat": 130.0, "pitch": 36, "len": 3, "vel": 127},
{"f": 1675, "beat": 134.0, "pitch": 36, "len": 3, "vel": 127},
{"f": 1725, "beat": 138.0, "pitch": 36, "len": 3, "vel": 127}
]
}
}
At 144 BPM a beat is 12.5 frames, so a note between frames goes to the nearest one: beat 130.5 lands on frame 1631. At tempos where a beat is a whole number of frames (150 BPM at 30 fps is 12) every note lands exactly.
Choose Suno if
- The song needs to be sung.
- You want a polished, genre-typical song from one sentence and won't change it much.
- You don't want to install anything.
Choose ismail if
- The music is instrumental: a score, a bed under narration, a loop, a beat, sound design.
- You want to change the music the way you change the visuals: one bar at a time.
- You want the visuals to land on the music exactly, from the notes and not from guessing.
- You want the files on your disk with no download limit, and the right to use them commercially.
- You want one agent session to make both the picture and the sound.
How it goes in practice
- Paste this to the agent you already use for the video:
Set me up with ismail: https://github.com/newsbubbles/ismail
- Tell it the length, tempo and feel, and where the big moments are ("drop at 0:42, quiet under the voice-over from 1:10").
- It writes the parts, renders, and reads the render back as text (levels per bar, drum lanes, structure) to catch problems before you listen.
- You listen and say what to change. It changes that part, not the whole track.
- Your video code reads the event file and cuts on the beat.
Where ismail falls short
From ismail's own README: a first sketch from a blank page can sound generic; it is strongest when there is something to measure. No singing voice, one tempo per project, no MIDI import or export, no plugin hosting. It needs a capable agent, and your ear is the final judge.
Sources, read 19 October 2026 and re-checked 22 October 2026: suno.com/pricing, suno.com/terms (effective 3 September 2026), help.suno.com article 13926209 (download limits), suno.com/release-notes. ismail facts from its README, docs/reference.md and the ismail.video source.