The tools changed this month. Whether artists want them is a different question.
Suno released version 5.5 with voice cloning that requires identity verification. Midjourney V7 added a draft mode that cuts iteration costs in half. ACE-Step 1.5 now runs commercial-grade music generation on consumer GPUs. And two new surveys—one from Artsy on galleries, one from the Authors Guild on writers—paint a picture of an industry that’s adopting AI for operations while remaining skeptical about AI as a creative medium.
Here’s what actually changed, who’s using it, and what they think.
Music: Suno 5.5 and the Voice Verification Play
Suno 5.5 launched March 26 with three headline features: Voices, Custom Models, and My Taste.
Voices lets subscribers sing AI-generated songs in their own voice. The catch: Suno requires live verification. You record a random phrase through the app, and the system matches it against your uploaded vocal samples. Voices remain private—only the uploader can use them. This is Suno’s answer to the voice cloning consent problem.
Custom Models let Pro and Premier subscribers train personalized versions of 5.5 on their own creations. You can build up to three models that learn your style. This matters for artists developing a consistent sound—the model adapts to you rather than you adapting to it.
My Taste learns your genre and mood preferences over time. It’s available to free users and functions like a recommendation layer for generation.
The business angle: Suno announced it’s “opening a new chapter in partnership with artists and the music business,” positioning these features as the foundation for label-licensed models coming later this year.
The Local Alternative: ACE-Step 1.5
If you’d rather not train Suno’s models with your voice, ACE-Step 1.5 offers a different path.
Released January 28, ACE-Step is a joint project from ACE Studio and StepFun. It generates full songs in under 10 seconds on an RTX 3090 with less than 4GB VRAM. On the SongEval benchmark, it outperforms Suno v5.
Key capabilities:
- Text-to-music from 10 seconds to 10 minutes
- Cover generation (reinterpret tracks in new styles)
- Audio repainting (regenerate specific sections)
- Vocal-to-BGM (generate accompaniment from vocals)
- LoRA fine-tuning for personalized styles
The license matters: ACE-Step is MIT-licensed, trained on legally licensed and royalty-free music, and runs entirely locally. You keep full control over outputs. No consent verification required—because you’re not training a company’s model.
Udio’s Pivot to Licensed Remixes
Udio is going a different direction. After settling lawsuits with Universal Music Group (October 2025) and Warner Music Group, the platform disabled its download functionality and is rebuilding around licensed music.
The new model: Udio will let subscribers create remixes and mash-ups of pre-existing songs—but only from artists who opt in, and generated music must stay on the platform. It’s a constrained vision, but one that might actually survive the copyright courts.
Downloads remain disabled as of March 2026 during the licensing transition.
Visual Art: Midjourney V7 and the Draft Mode Trade-off
Midjourney V7 became the default model this month with several features aimed at professional workflows.
Draft Mode generates images at 10x speed for half the cost. Quality is lower, but aesthetics stay consistent with full renders. The workflow: iterate rapidly in draft, then enhance the ones you like to full quality. For concept artists and designers doing dozens of variations, this cuts costs significantly.
Voice Mode lets you describe images through speech. The model interprets natural language and generates as you talk.
Raw Mode disables Midjourney’s built-in aesthetic processing. The result: outputs that look more photographic and less stylized. Useful for photographers who want AI-assisted compositing without the signature Midjourney look.
Personalization profiles learn your aesthetic preferences over time, similar to Suno’s My Taste feature.
Adobe’s Licensing Advantage
Adobe announced a multi-year partnership with Runway, bringing Runway’s Gen-4.5 video model to Firefly. The differentiator: Adobe trained Firefly exclusively on Adobe Stock, openly licensed content, and public domain material.
For commercial work where copyright provenance matters, Adobe’s approach offers clearer legal standing than tools trained on scraped internet data.
Writing: Sudowrite’s Specialized Approach
In AI writing tools, the market has split between general-purpose chatbots (ChatGPT, Claude) and specialized fiction platforms (Sudowrite, NovelCrafter).
Sudowrite’s Muse model was trained specifically for narrative prose. Combined with Story Bible organization and style-matching features, it targets fiction writers who need continuity enforcement across long manuscripts.
A 2024 analysis of over 500 indie authors found that writers using both a general chatbot and a specialized fiction tool completed manuscripts 40% faster than those using a single tool—while maintaining equivalent reader satisfaction scores.
What Creatives Actually Think
The surveys this month tell a consistent story: adoption for operations, skepticism toward AI as medium.
The Artsy Gallery Survey
Artsy’s February 2026 survey of 300+ galleries found:
- AI is “rapidly becoming a practical tool for gallery infrastructure”—communications, research, planning, administration
- 28% of galleries have no formal definition of what counts as “AI art”
- The vast majority of gallery artists are not using AI in their practice
- Many artists are “openly critical of its ethical implications and impact on authorship”
- Collector interest in AI-generated work “remains limited”
The conclusion: AI’s short-term impact on commercial art will be “operational, driving how the art world works, rather than shaping the art itself.”
The Authors Guild Survey
The Authors Guild’s survey found that 67% of professional novelists now use AI writing tools, but with significant caveats:
- Only 42% of fiction authors use AI “at least sometimes”
- 82% worry about AI homogenizing literary style
- The most popular uses: brainstorming, search, finding the right words
- Only 11% use AI to create publishable text
- 90% believe authors should be compensated for AI training use
Writers use AI for ideation and editing. They don’t use it to write.
Visual Artists
Broader surveys of visual artists show persistent concerns:
- 74% of professional visual artists report lost income directly attributable to clients substituting AI-generated imagery for commissioned work
- 89% believe current copyright laws don’t protect them from generative AI
- 74% say scraping the internet for artwork is unethical
When asked if their opinion of artwork changes after learning it’s AI-generated, the overwhelming majority said yes—for the negative.
What This Means
The tool landscape is maturing. Suno’s verification, Adobe’s licensing, and Udio’s pivot all represent attempts to solve the consent and copyright problems that have defined AI creative tools since 2022.
But adoption remains discipline-specific:
- Musicians are the most active adopters, with 87% incorporating AI into at least part of their process
- Writers use AI for research and brainstorming, not drafting
- Visual artists remain the most skeptical, with significant income displacement concerns
The local AI alternative matters. ACE-Step, Voxtral, and self-hosted image generators offer artists tools without the data extraction trade-offs. Whether adoption shifts toward these options depends on how easy they become to run.
What You Can Try
Run ACE-Step locally:
git clone https://github.com/ace-step/ACE-Step-1.5
# Requires ~4GB VRAM, generates full songs in under 10 seconds on RTX 3090Suno 5.5 voice features require Pro or Premier subscription ($30/month or $60/month). Verification is mandatory for voice cloning.
Midjourney Draft Mode works through the web interface or Discord. Use --draft to generate at half cost.
Sudowrite offers a free trial for fiction writers. NovelCrafter provides similar features with different workflow assumptions.
The tools exist. Whether they fit your practice is a different question—and based on these surveys, most artists are still deciding.