Suno v3.5 Update: Improved Quality, Common Issues, and Pro Tips

The release of Suno v3.5 has generated a lot of excitement, promising significant upgrades in audio quality. While the update delivers on some fronts, the community has also reported a mixed bag of results, from stunningly clear tracks to frustratingly off-genre generations.

Let's dive into what's new with v3.5, explore the common issues users are facing, and share some pro tips to help you get the best possible results.

The Good: What v3.5 Does Well

Enhanced Audio Fidelity

Right out of the gate, many users have noticed a tangible improvement in overall sound quality. The instrumentation is often clearer and more defined, especially in certain genres.

"I made a Phonk Polka song and it sounds MUCH cleaner."

This is particularly true for electronic styles like dubstep and phonk, where the cleaner mix allows complex layers to shine. Check out this example to hear the improved clarity:

Example Song: Phonk Polka

Longer Generation Times (For a Reason)

You might notice that v3.5 takes a bit longer to generate songs. Most users see this as a worthy trade-off for higher-quality output.

"Not going to complain about that, as long as the quality is better as well."

The Challenges: Common v3.5 Issues

Despite the improvements, v3.5 has introduced some new quirks. Here are the most common problems users are encountering.

1. The "Pop-Washing" Effect

A frequent complaint is that v3.5 has a strong bias towards a generic pop sound, regardless of the prompt. Users requesting niche genres or heavier styles often find their creations infused with an unwelcome poppy feel.

"Trying to make metal and it sounds great, then the vocals come in and it is so poppy sounding every time."

Similarly, prompts for obscure or complex genre fusions may be ignored entirely.

"I reused my prompt of Indian folk, ambient, baccata mix, but got slow country music instead."

2. Persistent Vocal Artifacts

While the instrumental quality has improved, vocal quality remains a sticking point. The infamous “demon choir” effect and other distortions can still appear, detracting from an otherwise clean track.

"Anyone know if 3.5 or v4 cleared up the demon chorus/distorted vocals?"

This issue can undermine the improvements in instrumentation, as heard in this example where the vocals have a noticeable distorted quality:

Example Song with Vocal Artifacts

3. Prompt and Style Glitches

Some users have reported issues with the prompt generator, where it seems to forget or ignore parts of the input. This can lead to unexpected and incorrect song generations. A peculiar but effective workaround has been discovered:

"Put a full stop in the prompt when trying again and it generated fine."

Adding punctuation like a period at the end of your style or lyric prompt can sometimes force the model to re-evaluate the input and adhere to it more closely.

Pro Tips for Better Results with v3.5

Navigating these challenges requires a bit of strategy. Here are a few tips to help you steer v3.5 in the right direction:

1. Be Hyper-Specific in Your Prompts: Don't just rely on the style box. Reinforce your desired genre and instrumentation within the lyrics prompt itself using bracketed descriptions, like [Verse 1: slow, atmospheric black metal with blast beats]. This redundancy can help fight the model's pop bias.

2. Experiment with Punctuation: If the generator seems to be ignoring your prompt, try adding a period or other punctuation at the end. This simple trick can sometimes reset its focus and lead to a more accurate generation.

3. Dispute Off-Target Generations: If a song is completely unrelated to your prompt (e.g., you asked for metal and got pop), don't hesitate to use the dispute feature. This provides valuable feedback to the Suno team for future improvements.