Prompt length limits across Firefly and Sume
Adobe Firefly now takes longer prompts for Image and Video on the web. Sume documents a 5000-character cap for music and no stated cap elsewhere.

Sume documents one hard prompt cap: music prompts take 1 to 5000 characters. Image prompts must be a non-empty string and the video docs list prompt as a required string; neither page states a character limit. Adobe's September 2026 Firefly update says longer prompts are now allowed for Image and Video on the web.
What Firefly changed
Adobe's announcement says Firefly now supports a longer prompt length for Image and Video on web, and that credit usage can drop when you fit a complete idea into one prompt instead of iterating through several short ones. The post as read does not give a character count for the new limit, so this article does not quote one.
What Sume documents per model family
The table lists only what the Sume docs state. A blank means the docs name no limit, not that no limit exists.
| Family | Prompt rule | Source page |
|---|---|---|
| Music 1.0 | Required, 1 to 5000 characters | models/music |
| Image 1.0 | Required, non-empty string | models/image |
| Video generation | Required string, no limit stated | models/videos |
Length and cost on Sume music
A long music prompt does not cost more. The music docs state a fixed $0.125 per accepted generation, and the price does not vary by prompt length or optional image conditioning. That makes a detailed brief free of extra cost, which is the opposite of the incentive Firefly describes for credits.
The music docs also recommend a structured brief: emotion, genre, tempo as a number, key, two to four instruments, an arc with one named moment, and era. Seven short clauses fit well inside 5000 characters.
Why one complete prompt usually wins
Longer prompts help when a scene has several constraints: camera move, lighting, wardrobe, product placement and mood. Splitting those across three short prompts and three generations multiplies cost on any per-generation model, so it is worth writing the complete brief once, as Adobe suggests for credits.
The same advice has a limit. Models can ignore details that sit deep in a long prompt, and the Sume docs describe music direction axes as creative directions, not guaranteed settings. Verify the output instead of assuming that every clause was honored.
A practical rule for clients
Do not rely on a limit that is not documented. For image and video, validate length on your side against a number you choose, and handle a 400 response from the API if a request is rejected. For music, reject anything over 5000 characters before you submit.
- Music: enforce 5000 characters client-side
- Image and video: keep prompts to what the scene needs and handle 400 errors
- Re-read the model page when a vendor announces a longer limit
Sources
Related posts
More in Developers
- Vendors swap GPUs; keep one video job shape
Luma said on Jul 23, 2026 it runs video-to-video inference on AMD and Tensorwave. Your client should not care: one Sume job shape covers every model.
- Animate a still by API: first-frame jobs on Sume
fal lists FLUX 3 as animating one still into video. On Sume you send the still as a first frame to a video model; this Python script submits and polls.
- Ask for the aspect ratio first: an MCP input_required round trip
MCP multi round-trip requests let a tool answer input_required to ask for an aspect ratio or spend approval before a render, with state in requestState.
- Attach a terminal to a run: ant sessions connect vs Sume jobs
The ant CLI can attach to a Managed Agents session. For Sume generation jobs, use sume jobs watch and MCP jobs_wait instead, and never resubmit a paid job.
Written by Sume